If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

In this issue: AstraZeneca joins a shared 1,152-GPU supercomputer in Sweden as pharma's compute race outpaces its published results — Anthropic's S-1 shows 9 of the 10 largest pharma companies are paying customers — insitro and AI Proteins sign deals worth $1.5B and $400M+ — Iambic prices its IPO while Recursion raises $200M — OpenAI buys Orchestra Bio and sets nearly 2,000 agents loose on a brain atlas — China drafts a ban on foreign takeovers of its CGT and siRNA platforms, and more…

This issue is presented by: Scala Biodesign

Antibodies by design: fix developability and affinity from the first design round

Live webinar · Wednesday, November 4, 2026 · 10:00 AM ET | 4:00 PM CET | 5:00 PM Israel time

Scala Biodesign combines AI with physics-based calculations to design antibody frameworks and antigen-binding regions, building on a decade of development across antibodies, vaccines, gene therapies and other biotherapeutics.

In this session, co-founders Dr. Ravit Netzer (CEO) and Dr. Adi Goldenzweig (CTO) draw on Scala’s work with top-10 pharma companies and biotechs to walk through experimental antibody case studies: the starting problem, the sequence changes designed, and how variants compare with the parent antibody across affinity and developability assays — which properties improved and which were retained.

You’ll learn how Scala’s design framework addresses:

• Stability: aggregation, self-association, polyreactivity and expression yields

• Activity: affinity and selectivity

• Program constraints: controlling the location and load of mutations

• Sequence diversity: antibodies with diverse sequences and V-gene combinations that meet binding and developability requirements, giving teams multiple candidates to advance

• Antigens: an antigen design module to improve the expression and stability of antigens and drug targets

Teams developing scFvs, VHHs, multispecifics or ADCs can use the session to assess which design approaches are worth testing in their own programs.

👉 Register for the webinar

Main Story

On October 15, AstraZeneca and a group of Swedish industrials, including Ericsson, Saab, SEB and Wallenberg Investments, launched Sferical One, a shared NVIDIA supercomputer reportedly built on 1,152 Blackwell Ultra GPUs for more than SEK 1 billion. AstraZeneca plans to use it for drug discovery.

The list goes on. Lilly’s LillyPod has been live since February with 1,016 GPUs, Roche now runs more than 3,500, BMS’s Vera Rubin system is due in Q1 2027, and Novo Nordisk and Zealand Pharma compute on Denmark’s Gefion. The software is following: Boltz will run its models inside BMS’s SuperPOD, trained on BMS data.

What’s still missing is the output. None of these systems has been tied to a published molecule or utilization figure yet, and each “most powerful” claim uses a different yardstick. Lilly’s chief AI officer Thomas Fuchs put it plainly: the bottleneck is biology.

More compute also means more candidates to check. Google’s new Deep Review agent has already verified more than 16,000 manuscripts at major ML conferences, and in drug discovery, Moe Alsumidaie argues the bottleneck has simply moved to clinical validation.

So what should count as a return on a SuperPOD?

🔒 From our members' deep dives: Big Pharma Is Betting On LLMs and Agents. What Would Count As Success? — the deals, and how to measure each kind of value.

The Brief

💰 Follow the Money

• A frontier lab files, with pharma on the customer list — Anthropic filed for a Nasdaq IPO under “ANT” today, and its S-1 shows 9 of the 10 largest Fortune 500 pharma companies each spent at least $100,000 with it in the year to Sept 30. The filing lists healthcare and life sciences first among its new markets and names Claude Science as a core product; first-half 2026 revenue was $16.3B, the FT reported.

• AI drug discovery tests the public markets — Iambic priced its IPO at $16 a share for $150M on Oct 15, the same day Recursion raised $200M in a follow-on at $3.25 that it says funds operations into 2029. We track the public cohort in 13 AI Drug Discovery Companies to Watch in 2026.

• Pricing a platform before the clinic — Isomorphic Labs is reportedly in early talks to raise at a $40B–$50B valuation, Bloomberg reported, after a $2.1B Series B in May.

• Robots instead of CROs — Warp Sciences, co-founded by former OpenAI science lead Kevin Weil, launched with $215M led by Lightspeed to automate molecule making and testing in the US, work now often sent to Chinese CROs.

• Engineered working dogs — Colossal-backed Warhound raised a $200M seed at a $1.2B valuation to clone and gene-edit military dogs equipped with Anduril gear; first births are expected in early 2027.

• Chemistry-native AI — Oakland-based Albert Invent raised a $55M Series B led by TCV for its R&D platform, used by BASF, Henkel and Clorox to retrieve lab knowledge and design formulations.

• Patient data from four countries — BioStack raised a $9.5M seed co-led by Peak XV and Afore Capital to build de-identified longitudinal datasets from a claimed 100M+ records in the US, India, Brazil and Argentina.

  ---

🤝 Deals & Alliances

• Cardiometabolic targets for the Virtual Human — AstraZeneca will pay insitro $40M upfront and up to ~$1.46B in milestones, plus royalties, to take targets in MASH and cardiomyopathies to development candidate.

• Miniproteins by design — Astellas licensed AI Proteins’ de novo miniprotein platform in a deal worth more than $400M including milestones.

• A milestone and a trial predictor — Insilico hit the first milestone in its $888M oncology deal with Servier and launched inClinico Agent, which reportedly predicts trial efficacy with an AUROC of 0.828 (medRxiv preprint).

• Open models, inside the firewall — Boltz will deploy Boltz Lab inside BMS’s NVIDIA SuperPOD and co-develop custom models on BMS data; no terms disclosed.

• Exploratory, for now — Oracle and AstraZeneca signed a non-binding agreement to explore AI and real-world data for trial recruitment and evidence generation.

  Share

  ---

🔬 Science to Watch

• Genetics without the lab test — Sooknah, Xu et al. (medRxiv preprint, Apr 14, 2026) at Calico found disease genes directly from AI embeddings of MRI scans, linking a CASP9 variant to kidney failure and an ADGRG6 variant to asthma and COPD, with no blood or breathing tests.

• Mapping the nervous system — Precision Neuroscience released a neural atlas from 70 patients and more than 180 hours of recordings with its thin-film brain interface.

• Dodo, 95% done — Colossal says its dodo genome is about 95% complete, with ~2,300 candidate gene-editing targets for beak shape still being validated in its closest living relative, the Nicobar pigeon.

• Your brain as a password — Paris-based Yneuro, backed by Bpifrance, is building Neuro ID, software that continuously verifies identity from EEG signals in earbuds and smart glasses.

• Banking immune cells — Cellino’s Reservoir now cryopreserves people’s immune cells for $5,000 plus $270 a year, betting on future personal cell therapies.

  ---

📋 From the Regulators

• China moves to keep its platforms at home — China’s Ministry of Commerce proposed banning foreign acquisition of proprietary cell and gene therapy and siRNA platforms; out-licensing and CRO/CDMO services are excluded, and comments are due Nov 15.

  Leave a comment

A Thing To Know This Week

OpenAI goes full “bio” mode

Nearly 2,000 GPT-Rosalind agents, built by PerturbAI and OpenAI with Scripps Research, analyzed a whole-brain CRISPR atlas of the mouse in 32 minutes — work estimated to take humans months. In blinded comparisons, experts preferred the agents’ interpretations 75% of the time (bioRxiv preprint).

A day earlier, OpenAI acquired Orchestra Bio, whose agents monitor drug programs and flag problems such as a failed manufacturing batch, and folded it into its Rosalind life sciences offering. Put together, OpenAI is building agents for both ends of the pipeline: finding targets upstream and running programs downstream.

Xaira, too, introduced X-Scientist this week, reporting hit rates on par with human-curated picks in a fraction of the time.

Usual caveats apply: preprints, self-reported results, and expert preference is just that — preference. Marketing play or a serious push into biotech?

Strategic Signals…

According to Arjun Murthy of Maven Bio, every bispecific ADC at Phase 3 or later was discovered in China

His team analyzed 243 bispecific and biparatopic ADC programs across 114 developers. Most are still preclinical, but China originated 46 of the 53 clinical programs and all four at Phase 3 or later, while six Western-owned clinical programs have been discontinued since 2018.

The timing matters: several of those Chinese assets, including BMS-partnered iza-bren and CTTQ’s TQB2102, reported Phase 3 data at ESMO in Madrid this week. And 36 of 46 Chinese clinical assets still have no disclosed ex-China deal; those that did, like iza-bren ($800M upfront from BMS) and JSKN016 ($125M upfront from Pathos AI), have set the price.

🧬 Time will tell, but with targets and payloads crowding (23 EGFR × MET programs, and topoisomerase I payloads in at least 77%), Murthy expects follow-on programs to compete on tolerability, dosing and patient selection rather than novelty. Combined with China’s draft limits on selling its platforms, the West’s role looks increasingly like licensing the assets, not owning where they come from.

Want the deep dives? Members get the full analysis of every deep dive plus our complete archive. Become a member →

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii and WTMB Team!

👉 Register for the webinar

Read also:

Biotech IPOs Roar Back, With Nvidia-Backed Iambic Next in Line🔒

The $2.7B Question: How Do We Know If Isomorphic’s AI Works?

Big Pharma Is Betting On LLMs and Agents. What Would Count As Success?

One recent sign: in October 2026, Iambic raised $150 million in a Nasdaq IPO, saying in its filing that its AI approach can “disrupt legacy drug discovery paradigms.” Investors and pharma partners increasingly judge companies like Iambic on clinical data rather than on technology alone.

Since the previous update of BiopharmaTrend company tracker back in 2024, the evidence has cut both ways. Generate:Biomedicines raised $400 million on Nasdaq, and Insilico Medicine completed Hong Kong’s largest biotech IPO of 2025. Insilico’s rentosertib became the first generative-AI-discovered drug to enter Phase 3, and Generate’s lead antibody is in pivotal asthma trials. Meanwhile, there are setbacks, including Neumora’s lead drug failed two Phase 3 studies, a reminder that machine-learning/datascience driven discovery does not remove clinical risk.

These companies are a small slice of a much larger field. According to BiopharmaTrend’s internal database, more than 500 startups and scaleups use machine learning and generative AI in drug discovery, from point tools for target finding or drug repurposing to end-to-end platforms that run from hypothesis to clinical trial design. In earlier work we reviewed a decade of AI implementation and the novelty of targets pursued by AI-discovered molecules. Our conclusion then still holds: the field is early, with setbacks and inflated expectations alongside real gains.

In a new article update, we profile 13 illustrative AI-enabled drug discovery/development players, most of them clinical-stage and publicly traded, with the latest milestones for each.

Disclaimer: This post is not investment advice of any kind; it is for informational purposes only. Neither WTMB Research & Media, S.L. nor the authors of this post endorse or recommend any particular stocks or investment opportunities. Consult relevant professional service providers for any investment decisions at all times.

Keep Reading

Quick hits from across the techbio space:

AI agents built by PerturbAI and OpenAI analyzed a whole-brain genomic dataset and synthesized insights in 32 minutes.🧠 read here.

Astellas links up with AI Proteins in $400M computational drug discovery deal

Eli Lilly and Company commits up to $1.1B to tackle extrahepatic delivery, partnering with Hong Kong-listed Metis TechBio on tissue-specific LNPs. read here.

Albert Invent raised $55M in Series B funding for chemistry-focused AI platform to be used in R&D work. read here.

Xaira Therapeutics introduced agentic AI platform for drug discovery, bringing together frontier language models, its biological models and datasets, and adaptive research workflows. read here.

TechBio 3.0 Is Closing the Loop. Is “Prompt to Drug” Next?

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

In this issue: a new review names "TechBio 3.0" just as Alex Zhavoronkov says prompt-to-drug may be here soon for low-to-moderate novelty — Schrödinger's Bunsen now launches ~80% of the company's own modeling jobs — ByteDance spins out Anew Labs at a RMB 10B valuation — Tempus becomes part of the manufacturing chain for Merck and Moderna's personalized cancer vaccine — a scaled-back version of a Nobel-winning reprogramming trick shows first signs in human eyes, and more…

Main Story

A couple of days ago, Alex Zhavoronkov, founder and CEO of Insilico Medicine, posted that he thinks we are approaching a “pharmaceutical singularity”, and that “prompt to drug is now possible in low-moderate novelty space.” The real moat now, he added, is operational efficiency and the ability to deliver.

It is a strong statement, I’ll come back to it.

Now, last week, Krish Ramadurai and Abhirup Banerjee published a review in npj Drug Discovery that gives the underlying trend a name: TechBio 3.0. They split the last decade into three generations:

• TechBio 1.0 (≈2015–2019), prediction: ML models ranking and filtering compounds inside conventional pipelines.

• TechBio 2.0 (≈2020–2023), acceleration: generative models and end-to-end platforms, with Insilico’s IPF program going from target to preclinical candidate in about 18 months versus the usual 3–5 years.

• TechBio 3.0, the closed loop: multimodal models, generative design with toxicity and PK constraints built into the objective rather than filtered afterwards, and automated experiments feeding structured data straight back into the models.

The authors are fairly sober about the evidence. As of mid-2026, no AI-designed small molecule has FDA marketing approval. Rentosertib remains the most advanced proof point, with Phase IIa data in 71 IPF patients (a mean +98.4 mL in forced vital capacity at the top dose) and Phase III expected in the second half of 2026. They also cite Iambic, which filed for its Nasdaq IPO a few weeks ago, as a working example of weekly design-make-test cycles.

It is interesting how much of the news lately is about the loop rather than the models:

• Lilly’s TuneLab released a nanobody developability model trained on 2,500+ VHHs that BigHat characterized in six one-week design rounds.

• AstraZeneca and Carterra are co-building hardware and software to produce standardized binding data for lab-in-the-loop biologics discovery.

• PacBio and TychoBio are sequencing full-length transcriptomes from 10,000+ samples to learn how antisense oligonucleotides behave across the transcriptome.

None of these is a model launch, but rather they are about producing data a model can trust (details in The Brief below).

The execution layer is moving too. Over the weekend, I published my interview with Robert Abel, CSO of Schrödinger, about Bunsen, the company’s agentic co-scientist. One number is especially telling: roughly 80% of the computational modeling jobs on Schrödinger’s own HPC (e.g. free energy calculations, molecular dynamics, quantum calculations, etc.) are now run by Bunsen. Importantly, the LLM is not predicting molecular properties itself. It chooses among validated physics methods and runs them, after a scientist has reviewed the work plan.

That brings me back to Alex’s qualifier. Abel makes a related point from the other side: ML accuracy generally worsens the further you move from the training data, and real discovery often happens exactly there — novel scaffolds, new resistance mutations, targets with few known ligands. His answer is physics as a data generator. A typical lead-optimization project might synthesize fewer than 1,000 designed molecules in a year, while atomistic simulations can produce several thousand data points in an evening.

So “prompt to drug in low-to-moderate novelty space” and “closed loops that work where data exists” may be describing the same frontier. The open question is high novelty, i.e., first-in-class targets, new modalities, where the loop has the least data to close on. The npj authors list it among the unresolved problems: predictors need applicability-domain checks before molecules go to synthesis. The authors are warning that a model can give a precise-looking score for a molecule it effectively knows nothing about. Their fix is to keep the model’s confidence level attached to each prediction at every step, so a shaky guess is not used for the final choice.

All in all, TechBio 3.0 may turn out to be less about a smarter model and more about who owns the data factory and the execution layer.

Read the full interview with Robert Abel here: https://www.biopharmatrend.com/interviews/interview-with-robert-abel-from-schrodinger/

🧠 Neurotech, nine months later

In January, we argued that 2025 was the year neurotech started designing around use cases: speech restoration as a defined clinical endpoint, noninvasive BCIs scaling on large datasets, organoids turning into standardized tools. This week offers two small tests of that thesis.

Silicon Valley-based Bridge Neurotech emerged from stealth with $13.5M from General Catalyst, reMind Capital, Sunflower Capital, Scott Sandell and others, to read brain activity through the skull with functional ultrasound imaging (fUSI) and AI decoding, with no surgery or implant. Founder Will Biederman designed the neural interface chip later used in Neuralink’s first implants during his Berkeley PhD, and co-founded Forest Neurotech. Bridge has licensed Butterfly Network’s Ultrasound-on-Chip through Butterfly Embedded and is enrolling its first research study in Redwood City.

What held: the pitch is a concrete one (a wearable, non-surgical interface, starting with a research study), not a general mind-reading device. What’s still open is what the company itself names as its two hurdles: getting reliable signal through the skull, and decoders that generalize across people. That’s the same cross-subject problem Tether Evo’s papers attacked in #85.

On brain delivery, WhiteLab Genomics’ AI-designed AAV capsids showed 50x higher brain DNA enrichment than AAV9 in IV-dosed mice (more in Follow the Money). It’s still mouse data, and larger-animal results will be the real check.

🔒 From our members’ deep dives: 2025 Neurotech Review: BCIs, Brain Delivery, Organoids and Neuro-AI Move Closer to Clinic — more than $1.3B in tracked financings led by Neuralink’s $650M round, the clinical proof points, and the forward signals we flagged for 2026. Worth rereading against what this year has delivered. Read it as a member →

The Brief

💰 Follow the Money

• Big tech spins out AI pharma — ByteDance’s Anew Labs raised a $290M angel round at a reported ~$1.5B valuation. It has four preclinical assets, led by an oral IL-17A/F inhibitor.

• mRNA made in cells — Sensible Biotechnologies raised $47M, including up to $20M in EU and Slovak grants, to scale cell-based mRNA manufacturing as an alternative to in vitro transcription.

• AI-designed capsids — WhiteLab Genomics raised a $26M Series B led by AVP. Its AAV capsids reached the brain at 50x the level of AAV9 in mice, with no detectable liver signal.

• Ultrasound BCI out of stealth — Bridge Neurotech raised $13.5M from General Catalyst and others to read brain activity through the skull with ultrasound and AI, no implant needed.

• Watching immune cells kill — SpectaCell closed a seed round of EUR 6 million led by imec.xpand for a platform that tracks immunotherapy–target interactions in real time.

  ---

🤝 Deals & Alliances

• Sequencing as a manufacturing step — Tempus expanded its deal with Moderna and Merck to run sample logistics and the sequencing used to design each patient’s personalized cancer vaccine, V940.

• Data for Lilly’s models — Eli Lilly’s TuneLab released NanoLab, a nanobody developability model trained on 2,500+ VHHs characterized by BigHat Biosciences.

• Lab data built for AI — Carterra and AstraZeneca will co-build hardware and software to produce standardized binding data for lab-in-the-loop biologics discovery.

• Long-read RNA for model training — PacBio and TychoBio will sequence 10,000+ samples to train models predicting how antisense therapies for rare diseases act across the transcriptome.

• Data first, agents next — Novo Nordisk runs ~900 TB of data on Snowflake for nearly 40,000 users and plans to turn its commercial data into an agentic system, per a Snowflake case study.

  Share

  ---

🔬 Science to Watch

• Neurotech, nine months on. In January we argued that 2025 was when neurotech started building devices around specific clinical jobs, with noninvasive BCIs advancing mainly through AI trained on large brain datasets. This week, Bridge Neurotech’s ultrasound BCI round fits that picture but still faces the hard part, decoders that work across people. Separately, on getting therapies into the brain, WhiteLab’s AI-designed capsids look promising but so far only in mice (both in Follow the Money).

  🔒 From our members’ deep dives: 2025 Neurotech Review: BCIs, Brain Delivery, Organoids and Neuro-AI Move Closer to Clinic — more than $1.3B in tracked financings led by Neuralink’s $650M round, the clinical proof points, and the forward signals we flagged for 2026. Useful as a baseline for judging what 2026 has actually delivered. Read it as a member →

• Partial reprogramming reaches human eyes — Life Biosciences reported early Phase 1 data for its three-factor epigenetic reprogramming therapy in glaucoma: no treatment-related side effects so far, and visual-field gains in two of three patients at eight weeks. It is a conference presentation in three people, so very early.

  ---

💁‍♀️ People 💁

• Miriam González, a software engineer in Spain with a rare metastatic breast cancer, built AI agents that helped her doctors choose a biopsy site. Anthropic has since taken her project into its AI for Science program.

Share

A Think To Know This Week

The 2026 Nobel Prize in Medicine goes to optogenetics

On Oct 5, the Nobel Assembly at Karolinska Institutet awarded the 2026 Nobel Prize in Physiology or Medicine to Georg Nagel (University of Würzburg), Peter Hegemann (Humboldt University Berlin) and Karl Deisseroth (Stanford) for their work on light-gated ion channels and optogenetics. The three share 12 million Swedish kronor, about $1.2M.

The idea is that some green algae use light-sensitive proteins, called channelrhodopsins, to sense light and swim toward it. In the early 2000s, Nagel and Hegemann showed these proteins work as channels that open when light hits them, letting charged particles flow into the cell. In 2005, Deisseroth’s lab put one of them into mammalian neurons and showed that a flash of light could make those neurons fire on command.

That turned light into a switch for specific brain cells. Researchers can now genetically target one cell type, turn it on or off in milliseconds, and watch what changes in behavior. The Nobel committee noted that the method is now used in labs worldwide to study how the brain works.

Specifically, optogenetics became a standard tool for working out which circuits drive pain, addiction, movement or mood, which is the kind of evidence drug and neuromodulation programs build on. It has also reached patients: in 2021, a man blinded by retinitis pigmentosa partially regained sight after an optogenetic gene therapy, reported in Nature Medicine.

Strategic Signals…

According to Padraig McDonnell, President and CEO of Agilent, digital pathology is becoming the foundation for pathology data generation, analysis and application

Agilent agreed to acquire Proscia in an all-cash deal: $240M at closing plus up to $30M in performance milestones. Proscia’s vendor-agnostic software serves as a central operating layer for digital pathology. It connects scanners, lab information systems, image analysis apps, AI tools, people and data for more than 4,000 labs, hospitals and clinics and 16 of the top 20 pharma companies. It has trailing-twelve-month revenue above $18M and FDA 510(k) clearance for primary diagnosis.

Time will tell, but an instruments-and-reagents company paying roughly 13x trailing revenue (about 15x if the milestones are paid) for a software layer says where it thinks value is moving. Agilent already sells the stains, and it closed the $950M Biocare acquisition in June. What it didn’t own, despite distributing Proscia’s software since 2022, was the layer between the slide and the model. McDonnell says the deal establishes an Agilent platform connecting data generation with AI-based interpretation, from biomarker research through routine clinical use.

Members get the full access to every deep dive plus our complete archive. Become a member →

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii and WTMB Team!

Read also:

Biotech IPOs Roar Back, With Nvidia-Backed Iambic Next in Line🔒

The $2.7B Question: How Do We Know If Isomorphic’s AI Works?

Big Pharma Is Betting On LLMs and Agents. What Would Count As Success?

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

In this issue: OpenAI's fund leads Biossil to unicorn status on a bet that failed drugs failed the wrong patients — Anthropic says it won't take its own programs into the clinic — Enveda, Anew Labs, Varda and Precision each raise $250M+ — Roche lays out a path to autonomous labs — optogenetics wins the Nobel — Xaira finally shows its antibody pipeline — a microbe built to live on Mars, and more…

Main Story

In April, Toronto-based Biossil emerged from stealth with $70M and a plan that sounded a bit contrarian for an AI company: instead of designing new molecules, buy or license the ones that failed in their own clinical trials.

About five months later, Biossil raised another $153M (C$216M) at a $1B valuation, this time led by the OpenAI Startup Fund, with Founders Fund, Quiet Capital, Modern Capital, Golden Ventures, Panache Ventures, Duke University’s endowment and the Abu Dhabi Investment Council participating. The money goes to expanding clinical development capacity and acquiring more drug candidates, with programs spanning sickle cell disease, idiopathic pulmonary fibrosis, glioblastoma, breast cancer and Alzheimer’s disease.

The idea is not classic drug repurposing. Repurposing looks for a new disease for an old molecule. Biossil’s approach is closer to asking whether the drug actually worked — just not for everyone in the trial. If an efficacy signal existed in a more precisely defined patient group, it may have been diluted across a broad trial population.

That reframes a familiar number. Pharma’s 90%+ clinical failure rate is usually read as a story about how hard biology is. Part of it, though, may be a story about the purely matchmaking side of things, and that is exactly the kind of problem where AI on rich clinical and molecular data could make a difference.

CEO Anthony Mouchantaf argues that AI-generated biomedical insights still have to be translated into the physical world: “That requires dedicated infrastructure capable of integrating AI and medicine together from discovery through development and into the clinic,” adding, “We’re building that infrastructure in the United States and Canada.”

The round also fits OpenAI’s broader push into life sciences. In April, it launched GPT-Rosalind, a reasoning model for biology, drug discovery and translational medicine, released as a research preview through its trusted access program with Amgen, Moderna, the Allen Institute and Thermo Fisher Scientific named as partners, on top of an earlier-announced drug discovery partnership with Novo Nordisk.

Biossil is not alone, though. We mapped the “drug rescue” category in our members’ deep dive — who else is working this angle (Lantern Pharma, Pathos AI, BPGbio, etc), what the Roivant precedent teaches, and the approaches that didn’t work.

🔒 Read the full deep dive → become a member — members get new deep dives plus the full archive.

The Brief

💰 Follow the Money

• Nature’s chemistry, AI-ranked — Enveda raised a $311M Series E at a $2B valuation, led by Catalio Capital Management (total raised: $845M+). Proceeds go to three clinical programs, including one for severe skin conditions and one for maintaining weight loss after stopping GLP-1s.

• ByteDance’s AI biotech prices in — Anew Labs raised $290M in its first external round at a reported $1.5B valuation, led by HSG, IDG Capital, GL-C Ventures/Hillhouse and 5Y Capital; ByteDance reportedly keeps 56%. Its work spans protein structure prediction, molecular design and antibodies, and its research includes small-molecule IL-17 inhibitors.

• Drug manufacturing in orbit — Varda raised a $251M Series D co-led by Lux Capital and Natural Capital at a $1.6B valuation. It has completed six capsule reentries and booked 28 launches for 2027–2029.

• BCI money keeps concentrating — Precision Neuroscience closed a $250M Series D co-led by Pershing Square and the Ackman Oxman Institute ($430M total). Its Layer 7 interface has been used in 100+ procedures across 18 institutions.

• From models to medicines — Basecamp Research raised $140M from investors including S32, NVIDIA and Menlo’s Anthology Fund to launch a pipeline of AI-designed therapeutics, starting with medicines that “reprogram the body to repair itself.”

• An AI-designed ADC enters the clinic — BigHat Biosciences raised a $75M Series C co-led by DFJ Growth and Premji Invest, mainly for BHB810, a CDH17-directed ADC for GI cancers that dosed its first Phase 1 patient in September.

  ---

🤝 Deals & Alliances

• Another Roche-group bet on AI discovery — Earendil Labs and Genentech signed a $1.5B research collaboration on cancer therapies, following Earendil’s autoimmune licensing deals with Sanofi.

• Paying for delivery, not the drug — Novartis paid $125M for Sironax’s platform for moving large molecules across the blood-brain barrier, its latest BBB deal after BioArctic and SciNeuro.

• Roche plans for lab autonomy — At Pharma Day, Roche laid out a path to autonomous labs, naming Medra and Anthropic as collaborators. It says 40% of recent pipeline decisions had a tracked AI contribution, and one program went from lead ID to clinical candidate in 18 months.

• AbbVie, argenx, Lundbeck and Takeda joined Ginkgo Datapoints and Apheris’s developability consortium: ~10,000 antibodies tested to one standard train a shared model that each member fine-tunes privately. First data due early 2027.

• AbbVie partnered with Valkai, starting in clinical sciences, as it sees pharma’s bottlenecks shifting from discovery to trials.

• Twist Bioscience joined Eli Lilly’s TuneLab as an antibody characterization data provider, after Ginkgo Datapoints.

• Boston-based Tremont AI emerged from stealth with Mass General Brigham-licensed foundation models and agents for drug safety pathology; funding not disclosed.

  Share

  ---

🔬 Science to Watch

• A Nobel for switching neurons with light — Karl Deisseroth, Peter Hegemann and Georg Nagel won the 2026 Nobel Prize in Physiology or Medicine for optogenetics, which uses light to turn specific brain cells on and off with millisecond precision.

• Antibodies “drug-ready from the start” — Xaira unveiled its X-Design platform and first programs in a company blog post: XA-1 (oncology) reached a progressable binder in seven weeks, and XA-4 hit a GPCR where conventional screens and llama immunization had failed.

• Pretraining on the brain — Neuralink pretrained each participant’s encoder on thousands of hours of their own unlabeled neural data (50,000+ hours across its trial), producing decoders that need far less recalibration — some users went from 10 minutes a day to 10 a week — and reports a BCI record of 11.32 bits per second.

• A microbe made for Mars — Pioneer Labs says it engineered a microbe that lives only on nutrients found on Mars and makes building materials.

• Measurement before models — Sculpta Bio says its new RNA barcoding and sequencing approach improves isoform capture at scale across bulk, single-cell and spatial applications.

  ---

💁‍♀️ People 💁

• Richard Bonneau left his role as Genentech’s global head of AI for drug discovery, Endpoints’ Andrew Dunn reported, the latest in a string of AI departures there including KyungHyun Cho, Stephen Ra and Nathan Frey; two sources cited tension with Aviv Regev, which Genentech denies.

• AMD will acquire World Labs, with Fei-Fei Li joining AMD as chief scientist.

• BrainGate founder John Donoghue joined Precision Neuroscience as Scientific Advisor.

Leave a comment

Strategic Signals…

According to Andrew Dunn’s report, Anthropic will not take its pipeline programs into human trials — and will not compete with its pharma customers

The quote, from Dunn’s report: “Anthropic does not plan to start human testing or Phase 1 trials for its pipeline programs, with Eric Kauderer-Abrams saying they will not compete with its pharma customers.”

Well, it is a big claim, and it changes a lot of assumptions about Anthropic’s current push into pharma, doesn’t it?

Time will tell, but it reads like a deliberate choice to stay a supplier rather than become a drug developer. And the customer list is growing fast: Bristol Myers Squibb put Claude in front of more than 30,000 employees earlier this year, Novo Nordisk signed its own deal (both covered in our last issue), and Roche named Anthropic as a collaborator on its path to autonomous labs at last week’s Pharma Day. If you are selling the reasoning layer into every big pharma’s R&D, owning clinical assets that compete with those same customers is an obvious conflict.

The open question is what “pipeline programs” Anthropic has at all, and what happens to them if they don’t go into humans — partnering or licensing would be the natural route, but the report, as quoted, doesn’t say.

A model retirement can be a regulatory event

Yujan Shrestha, MD, describes a problem few generative-AI device makers seem to plan for: Claude Sonnet 4 was released in May 2025 and retired in June 2026, about 13 months, shorter than most De Novo reviews, and one of his team’s API keys sent 10,236 failed requests in a single day afterwards.

Anthropic gives at least 60 days’ notice, less than the FDA’s 90-day goal for a Traditional 510(k), and swapping the model string can be a design change needing verification and validation. His advice: write model replacement into a PCCP (agreed with FDA in a Pre-Sub), keep a sequestered, labeled test set that can requalify a new model in days, and pin dated model snapshots rather than aliases that move on their own.

Members get the full access to every deep dive plus our complete archive. Become a member →

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii and WTMB Team!

Read also:

Biotech IPOs Roar Back, With Nvidia-Backed Iambic Next in Line🔒

The $2.7B Question: How Do We Know If Isomorphic’s AI Works?

Big Pharma Is Betting On LLMs and Agents. What Would Count As Success?

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

In this issue: pharma signs enterprise LLM deals at scale but it's unclear what would count as success — Iambic files for Nasdaq while Orbis lands Novo Nordisk for up to $1.4B — Mayo and Thermo Fisher start a million-biospecimen omics venture — Stanford turns papers into agents that talk to each other — the brain turns out to have grown from two separate roots all along, and more…

Main Story

Earlier this year, Bristol Myers Squibb put Claude in front of more than 30,000 employees, with a stated goal of halving the time from target selection to lead molecule.

Novo Nordisk's new Anthropic deal is the latest in the same trend. Besides, there is Merck that has committed up to $1 billion to Google Cloud, deploying Gemini Enterprise across R&D, manufacturing, and commercial. Amgen deployed ChatGPT Enterprise. Novo itself already had an OpenAI partnership from the spring... the list goes on.

Over the weekend, I finally completed my editorial article, Big Pharma Is Betting On LLMs and Agents. What Would Count As Success?, exploring this trend. The deals of this type carry two quite different value propositions, and the return on each needs quite different ways to measure.

One is the operational aspect: software engineering, regulatory and medical writing, knowledge retrieval, financial planning, etc. The evidence of potential value is real, albeit somewhat mixed, according to several studies outside pharma.

The other aspect is scientific reasoning, and it is probably harder to measure. While there are different studies and benchmarks, there is a notable new attempt at benchmarking, published in Cell, where authors assess the performance of several leading frontier models and their own smaller models against 17 aging biology tasks. Beyond presenting a framework for measuring AI in bio, this study also leads to a conclusion that the model size alone may not be the only key to what decides performance on biological data.

This is in line with a rapidly growing market of smaller but more specialized and more grounded AI platforms, also with LLM and agentic capabilities. Those include Schrödinger, OWKIN, Causaly, BenchSci, Insilico Medicine, Lantern Pharma, and others building on physics simulation, patient cohorts, or other curated knowledge graphs.

BenchSci, for example, makes the same argument from the demand side. Its EMET agentic environment was picked by argenx‘s scientists after their own competitive evaluation — which CEO Liran Belenzon called earning trust “at the bench, not in the boardroom”. argenx’s Tim Van Acker described it as “… a scientist in your pocket – always available, always reasoning across the evidence.”

Pharma is buying both camps, and in my article I list some deals and observations.

More important, though, is that an agentic layer on top of these tools lets people who were never trained on them take (some) part. For instance, Schrödinger‘s execs say Bunsen is for “drug hunters who are not computational chemists.”

Causaly, in its turn, is putting scientific reasoning into MS Teams for medical and commercial teams, allowing for more kinds of specialists in the same interdisciplinary work.

All in all, big pharma is increasingly betting on LLMs and agents, but the questions become more and more critical: how to assess success in a relevant way?

Read the full deep dive here: https://www.biopharmatrend.com/business-intelligence/big-pharma-is-betting-on-llms-and-agents-what-would-count-as-success/

Share

📅 I am excited to share that I’ll be attending BIOSPAIN 2026 as guest media.

BIOSPAIN is a massive event, recognized as the largest biotechnology conference organized by a national bioindustry association in Europe and one of the largest in the world for partnering and industry collaboration.

The Brief

💰 Follow the Money

• AI drug discovery goes public — San Diego-based Iambic filed for a Nasdaq IPO under "IAM," with J.P. Morgan, Jefferies, BofA Securities and Citigroup leading; share count and price are not yet set.

  Proceeds go mainly to IAM1363, an oral, brain-penetrant HER2 inhibitor in Phase 1 for HER2-altered solid tumors that the company claims is the only known HER2 TKI binding the inactive DFG-out conformation, with a registrational trial targeted "as early as 2027"; IAM217 (KIF18A) and IAM-C1 (CDK2/4) follow.

  The filing lands mid-rebound — after a trickle of listings in 2025, 2026 has already brought Aktis Oncology ($318M, January), Eikon Therapeutics ($381M, February), Generate:Biomedicines ($400M, the largest biotech IPO since 2024) and Kardigan ($400M upsized, June), with ADARx, TRexBio and Retension queued behind.

• AI for regulatory science — Paris-based Biolevate raises €30M Series A co-led by RAISE France and Orange Ventures, with MSD Global Health Innovation Fund, STATION F, EQT Ventures & Growth and Future4care participating, to scale its AI platform that turns scientific and regulatory knowledge into auditable, evidence-grade workflows for regulated life sciences. The company reports 20x ARR growth over 12 months and 12+ enterprise customers in production, including work with Sanofi, and will open a Boston office to drive US expansion.

• AI infrastructure for pharma — Mithrl raises a $20M Series A backed by Headline, AGI House and a group of pharma executives, to build custom AI infrastructure for biopharma R&D. The company is positioning at the middle layer of the AI-for-biology stack — proprietary data, biomedical world models and per-customer infrastructure that let each pharma team run frontier models on its own data and pipeline, rather than competing at the application layer.

• A verification layer for AI-generated biology — New York-based Polyphron raised a $20M seed to pair an automated "Tissue Foundry" growing human micro-tissue from stem cells with a "Tissue World Model" simulation layer, testing what AI-generated therapeutic hypotheses actually do in human biology. Former Google DeepMind Gemini researcher Vinh Q. Tran joins as co-founder and Chief AI Scientist.

  ---

🤝 Deals & Alliances

• Pre-symptomatic disease biology at scale — Mayo Clinic and Thermo Fisher launched Precure, LLC, a Mayo-majority JV generating multi-omics data from one million biospecimens linked to longitudinal clinical records, using Thermo Fisher's Olink Explore HT proteomics and Orbitrap mass spec. The aim is to detect molecular disease signals years before diagnosis; no deal value disclosed.

• Standardized data as the entry ticket — In a new agreement, Ginkgo Datapoints will supply small molecule ADME and antibody developability testing to companies in Lilly TuneLab, Eli Lilly's federated-learning platform that lets biotechs use models trained on hundreds of thousands of Lilly molecules without either side exposing proprietary data. Uniform assay protocols should help the shared models learn faster, and new members can submit Datapoints-generated data to unlock additional benefits; no financial terms disclosed.

• Oral macrocycles against injectable targets — Copenhagen-based Orbis Medicines signed a multi-target discovery and license deal with Novo Nordisk worth up to $1.4B in upfront and milestone payments plus tiered royalties, with Novo also taking an undisclosed equity stake. The collaboration will use Orbis's nGen platform, which pairs generative AI with automated chemistry, to develop orally bioavailable macrocycles for cardiometabolic targets.

  ---

🔬 Science to Watch

• The brain grows from two roots, not one — Jokhai, Dundes et al. (Nature Neuroscience, Sept 18, 2026) found that the thinking part of the brain and the brainstem that runs breathing and heartbeat come from two separate groups of cells that split apart at the very start of embryo development, rather than from one common pool as textbooks assume. Knowing this, the Stanford-led team finally grew human brainstem motor neurons in a dish — the cells lost in ALS and spinal muscular atrophy, which had resisted lab-growing until now.

• Papers that run themselves — Miao, Zou et al. (Nature, Sept 16, 2026) built Paper2Agent, which converts a manuscript's text, code and data into an MCP-based AI agent that reproduces the study and then answers questions or applies its methods to new datasets; 74 of 100 computational biology papers were successfully converted. In a paired demonstration, an AlphaGenome agent and an ADHD GWAS agent reportedly surfaced a previously unreported variant near MPHOSPH9 associated with ADHD risk.

  

  Share

  ---

📋 From the Regulators

• Pulling first-in-human trials back to the US — The FDA launched its Expedited IND Pilot on Sept 15 under HHS's Operation TrialBlazer, letting sponsors submit parts of an IND application as they're ready instead of waiting for the full package, with optional expert review partners and site setup running in parallel. Applications close Oct 30 and participants are picked Dec 18; the pilot prioritizes commercial INDs for novel Phase 1 trials of products with no prior clinical experience.

• A second AI organ model into FDA's qualification path — Absentia Labs’ co-founder and CEO, Farhan Khodaee, says the FDA accepted its AI model for predicting drug-induced cardiotoxicity into the ISTAND pathway, which qualifies novel drug development tools that don't fit existing FDA categories. It follows the company's Digital Liver model and forms the second organ system in a planned multi-organ AI safety platform aimed at replacing some animal toxicology work.

  ---

💁‍♀️ People 💁

• Cellular Intelligence, the rebranded Somite Therapeutics, building a universal virtual cell-signaling model to direct stem cells into therapeutic cell types, added Yann LeCun (AMI Labs, ex-Meta), Bob Langer (MIT, Moderna co-founder), Jens Nielsen (BioInnovation Institute) and Fabian Theis (Helmholtz Munich) to its Scientific Advisory Board, with Langer also joining the board as an observer.

Leave a comment

Strategic Signals…

According to Eric Lefkofsky, Founder/CEO at Tempus AI, healthcare will likely be the biggest beneficiary of AI

Time will tell, but it seems like Tempus AI is trying to position itself as an infrastructure player in the AI-centric future of life sciences already today. From recent notable updates, for example, the company is building a research platform of 100,000 whole genomes over the next several years. The longer-term goal is 1 million.

🧬 This idea is different from big genome programs like national biobanks, because the latter mostly sample the general population. They are not typically linked to how a disease progresses or how patients respond to treatment. Tempus’s dataset, in contrast, focuses on specific diseases and ties each genome to clinical outcomes over time. So, it is likely built for AI work, including training and fine-tuning models, not just as a database resource.

Also, Tempus is arguably moving beyond targeted gene panels (tests that read only selected genes) to reading the whole genome. The data joins its existing de-identified records: clinical histories, imaging, pathology and outcomes. Researchers could use it through Tempus Lens, so they can build and test models without moving data between systems. This plays in favor of usability, which naturally increases chances of better insights.

Lefkofsky says it himself: “A large dataset is only valuable if you can turn it into insight.”

Regarding the timing, this work has started, and early data is available through an Early Adopter Program, where more members join in waves, and full availability is planned for mid-2027.

On a personal note, I recently joined Rafael Rosengarten of Genialis' Talking Precision Medicine podcast (episode #61, Sept 17) to argue that counting "AI-discovered drugs" is the wrong yardstick — candidates come out of long collaborative processes where the AI contribution can't be cleanly separated, so R&D productivity, faster killing of weak candidates and actual software adoption tell you more.

We also got into why AI became so polarizing, what stays human as information gets commoditized, and why the next era may run on patient samples, organoids and whole-system models rather than single frontier models and target-ligand reductionism.

Listen here (around 50min): https://www.genialis.com/2026/09/17/the-kodak-moment-for-drug-discovery-ai-adaptation-and-what-stays-human-tpm-podcast-61/

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii and WTMB Team!

Read also:

Could Anthropic Disrupt the Techbio Companies Built for Pharma/Biotech?

What Do the New AI Model Releases by Anthropic and OpenAI Change for Life Sciences?

The Bigger Point Beyond Insilico’s Aging Research Milestone

Into the Dark: Finding Novel Drug Targets Within the Depths of Our Proteome

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

In this issue: a retired chemist designs a renin inhibitor with ChatGPT and Gemini — Astromech raises at $3.8B and Forus at $3B while BioXcel files Chapter 11 — Google maps the largest connectome yet — Claude runs a protein design campaign — and a stablecoin issuer quietly publishes three BCI papers.

Main story

Chemical & Engineering News (C&EN) recently published an interesting case study on using off-the-shelf AI chatbots to assist in drug design work.

As Bethany Halford explains in her article, at ACS Fall 2026 in Chicago, Guibai Liang, a veteran medicinal chemist who co-founded Sheo Pharmaceuticals after retiring from big pharma, presented a talk in the Division of Medicinal Chemistry about how his team designed a novel renin inhibitor for hypertension using ChatGPT and Gemini, off the shelf, with no proprietary model in the loop.

Renin is about as picked-over as targets get: every major pharma ran a program, and aliskiren (Novartis’s Tekturna/Rasilez) is the only one the FDA ever approved — a mediocre drug, in Liang’s words, mostly on bioavailability.

Sheo took Takeda’s TAK-272 as a starting point and put what Liang calls “critical and revealing questions” to the chatbots: what does the hydrogen bond donor on the piperidine do, is the morpholine oxygen needed for activity, is metabolic stability a concern.

The models answered by synthesizing decades of published structure-activity relationship (SAR) insights — “it’s impossible for human chemists to remember all those data to come up with a clear picture,” Liang said for C&EN. From those answers, the team designed a new chemical class, had a CRO make 200 compounds across several iterations, and landed on SHEO-054, which reportedly outperformed aliskiren in a monkey model of hypertension.

Caveats: The structure wasn’t disclosed, only a general scaffold, and the data are self-reported from a conference talk, no paper yet. Also, “outperformed aliskiren” in one primate model is a long way from a development candidate.

The AI here read the literature no human could hold in working memory and handed back a coherent SAR picture, and Liang himself flags that renin may be a special case precisely because so much was published before the field walked away. The enabling condition is a dense, abandoned literature, which describes a surprisingly large share of pharma’s dead targets, and points at a strategy that doesn’t need a $100M platform to run.

Liang’s takeaway was blunt: “If you are not using AI in your medicinal chemistry, start now.”

Read “This company designed a drug candidate with help from ChatGPT and Gemini“ by Bethany Halford.

Share

📅 I am excited to share that I’ll be a speaker at the upcoming TechBio Europe 2026 event in Paris, on November 30.

TechBio Europe 2026 brings Europe’s TechBio founders, investors, and pharma partners to Paris for one day on Nov 30 at Future4Care — financing, dealmaking, hiring at the bio/code intersection, plus a pitch contest for the continent’s newest startups.

TechBio Europe is organized by the TechBio Commission, co-led by WhiteLab Genomics and Scienta Lab, under the guidance of France Biotech, France Deeptech, Future4care, and proudly sponsored by In Extenso Innovation Croissance, ICOSA | European Intellectual Property Firm, BioLabs, McDermott Will & Schulte, BioTalent, and Wellcome Genome Campus.

TechBio Europe also kicks off Bioweek, a week of industry events in Paris.

Register for TechBio Europe and let’s meet: https://www.eventbrite.fr/e/techbio-europe-2026-tickets-1996434242562.

The Brief

🚀 New Launches

• Geodesic Intelligence launched out of the gate with two products, NovaDDE and NovaAtom-Lite-Preview, pitching an AI-native platform for protein therapeutics and describing its aim as "AGI for drug discovery" — no funding, team, validation data or partners disclosed in the announcement.

  ---

💰 Follow the Money

• Astromech, the Colossal Biosciences spinout founded by Ben Lamm and George Church, raised $20M led by Bob Nelsen (PEAK6, NeoGenesis, Builders VC and CAZ participating) at a company-reported $3.8B valuation ($60M total) to expand its genomic training datasets and model infrastructure for longevity work on evolutionary outliers like bowhead whales and Brandt's bats, though no externally validated performance, revenue or deployment data was disclosed.

• Paul Allen's science fund is backing AI BioDesign, a Seattle effort, with ~$95M over five years — $46.1M to the Allen Institute, $43.8M to UW, $4.7M to Fred Hutch — for a loop where AI designs proteins and genetic switches that don't exist in nature, labs test them, and the results retrain the models.

• BioXcel Therapeutics, the AI-driven neuroscience biotech behind the approved agitation drug Igalmi, filed for Chapter 11 with Teva as stalking-horse bidder at $57.5M upfront plus $67.5M in milestones — a cautionary marker for AI-derived pipelines

• Forus raised a $150M Series C at a $3B valuation led by Bain Capital Ventures, taking total equity past $300M for its AI-agent network that cuts prior authorization from weeks to under 48 hours across physicians, pharmacies, payers and manufacturers.

  ---

🤝 Deals & Alliances

• BenchSci and argenx partnered to apply BenchSci's agentic AI platform EMET to preclinical drug discovery, pitched by CEO Liran Belenzon and argenx's Tim Van Acker as an "always-on AI research partner."

• Owkin licensed K Pro, its agentic "AI scientist," plus multimodal patient data from its MOSAIC network to Servier for oncology research.

• Merck KGaA is paying PostEra "mid-double-digit millions" for its two fertility programs — oral small-molecule agonists of the FSH receptor and the LH/choriogonadotropin receptor that could replace the injection standard in IVF. Both are designed on PostEra's AI platform to solve the selectivity and half-life problems that sank earlier attempts, while PostEra keeps its PMOS program in-house with trials planned for early next year.

• Ginkgo Bioworks won a subcontract worth up to $17.5M under ARPA-H's GIVE program, joining Waterfall Scientific on ESCALATOR — a multi-organization effort to compress end-to-end RNA drug manufacturing onto a benchtop footprint.

  ---

🔬 Science to Watch

• Google Research and HHMI Janelia published in Cell the complete AI-reconstructed wiring diagram of the male fruit fly's brain and nerve cord — 166,000 neurons and 125 million synapses.

  

  It is the largest connectome to date, which, paired with the existing female map, lets researchers compare the sexes neuron by neuron and sets up the methods for mapping zebrafish and mouse brains next.

• AlphaGenome Atlas by Google DeepMind, scores all 9 billion possible mutations, coding and non-coding ones. They say that Gareth Hawkes, a Medical Research Council fellow at the University of Exeter, arguably found 22% more non-coding associations in the UK Biobank just by re-ranking with the Atlas.

• Anthropic reports Claude, orchestrating existing open-source design tools rather than modeling proteins itself, produced 354 wet-lab-confirmed binders from 1,320 designs across 15 targets — a 22.6–35.1% hit rate against the 10–15% industry norm. Though the work is unreviewed, has no human-expert control arm, and the best-performing models stay access-restricted on biosecurity grounds.

• Insilico Medicine Doses First Patient in GENESIS-IPF-3, the World's First Phase III Trial of a Generative AI-Driven Innovative Drug

  

  ---

📋 From the Regulators

• The FDA is putting sponsors on notice that it will refuse clinical data it cannot inspect or verify — with extra scrutiny on foreign Phase 1 and non-IND/IDE studies in countries where site access is denied or informed consent is questionable — and is expanding foreign inspections and publicizing where access was blocked.

  ---

💁‍♀️ People 💁

• Hon Weng Chong, founder of Cortical Labs, the Melbourne company behind the DishBrain work and the CL1 biological computer, announced on LinkedIn that he has resigned after seven years, saying he's proud of what was built and no longer has access to his company email.

  

Leave a comment

A Thing To Know This Week

Addressing one of the biggest challenges in brain-computer interfaces

Tether Evo, a frontier tech division of Tether, the stablecoin issuer, had three brain-computer interface (BCI) papers accepted at the Journal of Neural Engineering, Imaging Neuroscience, and Neural Networks, two of them with the University of Rome Tor Vergata.

All three attack the same bottleneck: every brain produces different signals, so decoders have historically been rebuilt from scratch for each patient, and the calibration burden is a real barrier to fielding these devices. The common move across the papers is an alignment step that maps different people’s neural data into a shared space, so one model can be trained across subjects and then fine-tuned for a new one.

The speech paper (preprint on bioRxiv) deals with a cross-subject neural-to-phoneme decoder trained on invasive recordings from multiple participants implanted in different cortical regions, which is the hard version of the problem. Tether reports it matches or beats single-patient systems while adapting to a new person in minutes to hours rather than the usual long calibration, and describes it as the first of its kind. The vision paper reconstructs what macaques were looking at from 200 ms of spiking data, identifying the exact image out of thousands 70% of the time.

The music paper ran fMRI on five people hearing 540 songs across 10 genres, hitting 61% genre identification against 10% chance and ~25% song identification among 60 candidates against under 2% — classical and jazz the most distinctive, metal and disco the most confusable.

It is notable that Tether took a $200M majority stake in Blackrock Neurotech in April 2024, the company behind a large share of the world's chronic human intracortical implants.

Share

Strategic Signals…

👉 AI labs are pivoting hard into health care to fix the notorious public perception trend. Anthropic is talking up biology ahead of its IPO, Nvidia and Lilly are building a $1B drug discovery lab, etc., partly because curing disease is a far better public story than data centers hogging power and driving up utility bills, and partly because life sciences diversify revenue as these companies head to public markets.

The catch is timing: real breakthroughs are years off, AI is making some health care more expensive, and biosecurity risk is real enough that Anthropic added safeguards to Claude Fable 5 after finding it could give a bad actor "significant uplift," so the goodwill is being spent well before the results arrive…

👉 AI-enabled drug repurposing is one of the most straightforward value propositions of the modern AI drug discovery movement. THPharm ran its Phase 3 metabolic drug THP-001 through Insilico Medicine’s PandaOmics (including 16 public datasets, 269 samples of post-treatment gene expression) to find new indications it might also treat, surfacing heart failure, fatty liver disease, and obesity as candidates.

The logic is capital efficiency: the drug already has human safety data, so each new indication starts with a big chunk of risk removed. Though these are computational hypotheses that still have to clear cell models, animals, and human trials.

👉 Aaron Blotnick argued on LinkedIn that OpenAI's integration of ChatGPT directly into Epic (the EHR holding records for 325M+ patients) just wiped out an entire startup category. Clinicians can now summarize labs, pull patient history, prep for visits and build clinical timelines without leaving the chart, plus a new plug-in pulling live data from ClinicalTrials.gov, CMS, RxNorm, DailyMed and PubMed. His conclusion: "If your startup's moat was 'AI + EHR integration,' it's time to find a new moat."

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii and WTMB Team!

Share

Read also:

What Do the New AI Model Releases by Anthropic and OpenAI Change for Life Sciences?

The Bigger Point Beyond Insilico’s Aging Research Milestone

Into the Dark: Finding Novel Drug Targets Within the Depths of Our Proteome

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

In this issue: The FDA opens the question of how you evaluate a clinical AI that answers differently every time — Network Bio, Harell Data and Lilly's TuneLab each try a different way of getting proprietary data into AI models without the owner handing it over — Claude runs a 48-hour protein design campaign on its own and publishes the prompt — Insilico Medicine sketches an ageing virtual cell — Cell publishes fifteen grand challenges for generative AI in biology — Nature Reviews Neurology asks what happens after a clinical model works — Faro raises $37M to speed up trials — GSK starts paying NOETIK to rent its models — An emerging market for DNA as a programmable building material. And more…

The FDA writes down how it might grade a clinical AI, and opens the docket

This month, the FDA’s Center for Devices and Radiological Health (CDRH) published Considerations for the Regulation of Generative AI-Enabled Medical Devices. It’s a discussion paper, not guidance, and comments close October 19.

It is a conceptual matter, with potentially far-reaching legal and business implications for anyone building AI-driven medical devices… so make sure to read it.

In a nutshell, the old approach to assessing medical devices is to assume a device answers the same way every time, i.e., like “good old software”. Generative AI does not behave this way, it can answer differently tomorrow, and the foundation model underneath it can be updated by a third party, influencing the device behavior in unpredictable ways. So, trying to evaluate the behavior of AI-driven devices precisely might be unrealistic. The paper’s own words: “… it may be unreasonable to evaluate every conceivable input…”

So how do you evaluate something like that?

The FDA’s tentative answer is that you do it roughly the way you evaluate a doctor. Exams to check they know the material and won’t do anything dangerous, then supervised practice, then keeping an eye on them once they are working alone.

In practice, for a device that could become benchmarking, then clinical confirmation — which might be retrospective review, or shadow deployment, or a full prospective trial depending on the stakes. Then monitoring for as long as it is in use. In other words, where there is no single right answer, the device gets compared to a panel of clinicians or a median clinician rather than to an ideal correct response.

How much scrutiny you get depends on what the software actually does (just informs, directs, or acts on its own), and how badly things can go if it is wrong. A simple diagram explains the framework:

An important food for thought for businesses building AI-driven medtech devices — about risk management. A lot of health apps give someone specific advice and then add "talk to your doctor" underneath, on the assumption that the disclaimer keeps them in the safe, informational category. The FDA is signalling that it probably does not… What does is what the output actually tells the person to do.

⚡ This issue is brought to you by SYDRA

The Brief

💰 Follow the Money

• In a historical milestone, Moderna and Merck's intismeran, an mRNA vaccine built individually for each patient using data of their own tumour's mutations, kept melanoma from returning for longer when added to Keytruda, in a Phase 3 of about 1,100 people whose tumours had been surgically removed. It's the first randomised Phase 3 win for a personalised cancer vaccine, though the actual numbers come later at a conference, and each dose still takes months to manufacture. Anyway, Moderna’s stock jumped 177% on the news, its biggest one-day move ever, with Merck also gaining substantial upward movement.

• Network Bio, a biobank-focused biotech startup, launched out of stealth with $50M and what it calls the world’s largest patient-tissue training dataset — 500,000+ patients with paired tissue, blood and longitudinal outcomes from Mass General Brigham, Penn, Duke and CU Anschutz on a common data model. Investors include Section 32, Thiel Bio, Founders Fund and Breyer Capital, which led the company’s first institutional round in 2022.

  Interestingly, Network Bio and NVIDIA are also building what they describe as the first foundation model trained on cell-free RNA, using Parabricks and BioNeMo Recipes. The company also disclosed a $30M+ co-development and licensing agreement with a Fortune 100 healthcare company to identify and interpret novel disease signatures using Network’s proprietary AI architecture.

• Harell Data, founded by Adaptive Biotechnologies co-founder Harlan Robins, raised $15M from Fuse and Cercano Management to let data owners host proprietary training sets and take a cut of the compute revenue from models trained on them, rather than waiting on downstream royalties.

  Why it matters: it prices scientific data as a rentable asset instead of an equity bet. Early datasets come from A-Alpha Bio and Adaptive.

• Faro, an AI startup working on clinical trials, raised $37M from Merck's venture fund and S32. CEO Scott Chetham says the startups chasing faster trial documents "just made tiny bits a little bit faster."

🤝 Deals & Alliances

• Eli Lilly’s TuneLab models are now in Benchling. AbLab predicts antibody developability (thermal stability, aggregation, viscosity, solubility, etc) from heavy and light chain sequences. ChemLab returns ADME/Tox predictions from a SMILES string.

  The catch: access is reciprocal. Participating biotechs contribute their own assay data through TuneLab’s federated learning platform, which never exposes the raw data, and admins install the models from Benchling’s Model Hub registry.

• NOETIK hit the first payment milestone in its five-year GSK deal by getting its OCTO-VC cancer foundation models running inside GSK's own systems, which matters mainly because GSK is paying subscription-style fees to rent the models rather than buying drug candidates, and that arrangement is now producing money.

🔬 Science to Watch

• Claude ran a 48-hour autonomous protein design campaign and published the prompt. Anthropic reports de novo binders against 14 of 15 targets — 1,320 designs, 354 of which bound in wet-lab testing at Adaptyv Bio and Twist Bioscience, with hit rates of 22.6–26.7% across the multi-target run and 35% in a single-target 24-hour campaign, against a stated field baseline of 10–15%.

  The caveat, per Endpoints News, noted by Andrew Dunn: Claude generated nothing much itself, it orchestrated open-source tools (PXDesign, RFdiffusion, BoltzGen) and reasoned over the results, and hit rate moves with target difficulty. Maltose-binding protein defeated all 90 designs against it. Designs, raw data, a 29-page technical report and the full ~16,000-word prompt are public.

• Insilico Medicine sketched out what an ageing virtual cell would look like. Most virtual cell models are trained on cells caught at a single moment, so they can tell you what a cell looks like but not how it got there or where it's going. Insilico wants to go further, making age the thing the model is operating on. Instead of one big gen network, they plan on using a group of AI agents that each handle a different level — molecules, cells, tissues, organs, the whole organism, communicating between each other.

  The idea is you could ask what happens if you, say, knock out a gene, and get the answer accounting for time and space. I think it is more of a concept for now, as I did not find specific data or paper yet. They say more thinking is coming at Aging Research and Drug Discovery Meeting in Boston soon.

  

• In a new article in Cell, authors published fifteen grand challenges for generative AI in cell biology, sorted into four categories: molecular interactions, molecular functions, cellular and systems function, and translation.

  Why it matters: the authors argue current systems are benchmarked on retrospective statistical tasks rather than decisions, and call for CASP-style held-out evaluation.

• A Nature Reviews Neurology perspective maps what happens after a clinical model works: outcome benefit, workflow fit, reimbursement and drift, arguing deployment is not the end of model development but another stage of evaluation.

📋 From the Regulators

• The FDA approved Revolution Medicines' daraxonrasib for previously treated metastatic pancreatic cancer — the first drug ever cleared that hits RAS broadly, the cancer gene that defeated the industry for forty years and drives nearly every pancreatic tumour. It doubled median survival to 13.2 months against 6.7 on chemotherapy in the Phase 3 RASolute 302 trial, and the FDA cleared it six and a half months ahead of schedule.

Leave a comment

⚡In partnership with SYDRA

Most AI drug-discovery companies begin with a target and use algorithms to optimize molecules around an assumption. SYDRA reverses the logic.

Our proprietary AI selects and generates novel chemistry; whole-organism lifespan assays reveal which molecules actually affect aging biology; and only the phenotype-proven winners advance into aged-mouse studies, human-cell target deconvolution and disease positioning.

Starting from six million molecules, we identified five significant lifespan hits, advanced two lead programs into late-life mouse pilots, and created a new generative chemistry cohort now entering synthesis. We don’t use AI to decorate a target hypothesis. We let living biology decide which AI discoveries deserve to become medicines.

SYDRA is now raising a CHF 2.6 million pre-seed round to accelerate lead programs toward pharma partnering.

A Thing To Know This Week

Beyond the Code: How DNA Nanotechnology is Moving from Molecular Origami to Programmable Biotech Platforms

In a new article published on BiopharmaTrend, Louise von Stechow and Marco Lolaico make the case that DNA is worth treating not only as a carrier of genetic information but as a construction material, one whose predictable base pairing lets researchers fold structures that place proteins, ligands or fluorophores exactly where they want them, and even build devices that switch shape or release cargo on cue.

In the article, Louise and Marco walk through where that has actually reached the market, which so far is mostly research tools and diagnostics:

- GATTAquant already sells DNA origami calibration standards for microscopy,

- Amplifold raised €5 million to make rapid tests roughly a hundred times more sensitive without changing the familiar lateral-flow format,

- Nanogami and MediQuant are building biochips and point-of-care drug monitoring, respectively… and so on.

Therapeutics are further back:

Plectonic's logic-gated T-cell engagers, DoriNano's cancer vaccine now working with Memorial Sloan Kettering and Daiichi Sankyo, DNA Nanobots' non-viral gene delivery — all still preclinical.

The authors are clear-eyed about why: the structures are expensive to make, they fall apart in the body, and nobody has settled how to prove batch-to-batch consistency.

Cost is improving fast, though (a 2026 study got to about $6 per milligram in the lab), but they argue cheaper DNA alone won't do it, and that the field needs standardized manufacturing, better stabilization, quantitative quality control, and regulators willing to say which structural attributes actually matter.

Read the article here.

Share

Strategic Signals…

Bart De Witte, co-founder and CEO of Isaree, argues that Apple's new M5 Ultra chip, 512GB of unified memory under a desk, kills the assumption that hospital AI has to mean shipping patient data to someone else's data centre, because a frontier-scale medical model now fits in local memory. His company is building for that world: an open platform where clinicians build their own agents, running on-device or on-prem, with GDPR compliance built into the architecture (especially relevant for EU-based facilities).

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii and WTMB Team!

Share

Read also:

Could Anthropic Disrupt the Techbio Companies Built for Pharma/Biotech? (Also: The Numbers Behind China’s Biopharma Ascent)

Is the Future of AI Drug Discovery Hybrid?

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

Cancer as a Data Problem: What AI Is Doing in Oncology

Where Tech Meets Bio, as well as BiopharmaTrend, are now published by WTMB Research & Media SL, a company based in Spain.

This follows the restructuring of our business. Our editorial commitment remains focused on covering the cutting-edge technological advances in pharma, biotech, and medtech industries, companies that drive those advances, and regular analysis of industry trends.

As part of this transition, we are updating our payment system. This means:

— If you are a free subscriber, no action is needed. Your subscription continues as normal.

— If you are a paid subscriber, your current subscription will be cancelled, and you will receive an automatic prorated refund for any unused portion of your subscription. We wish we could avoid it, but this is due to Substack's internal hard limitation to be able to transfer payment arrangements when the newsletter moves to another country.

We will notify you when the new payment system is live so you can resubscribe if you wish. We apologise for the inconvenience and deeply appreciate your support.

Importantly, the data controller for this newsletter is now WTMB Research & Media SL (CIF: B93869147), C/ Almirante Cadarso nº 26, 46005 Valencia, Spain. Please read our Privacy Policy to know your rights concerning your personal data.

You can unsubscribe at any time using the link at the bottom of this newsletter if you do not wish to receive future emails from us. However, we do hope you continue reading our work, because we are going to expand and deepen our analysis and coverage of the “tech+bio” frontiers.

Thank you for being part of our community!

In the meantime, keep reading this week’s highlights as we are gradually renewing our weekly cadence.

Cheers!

— Andrii and the Team of Where Tech Meets Bio newsletter, from sunny Spain🌴.

In this issue: a decade-in report card on AI drug discovery; frontier AI labs moving into the CRO and clinical-trial layer; agentic co-scientists landing inside pharma research organizations; what is happening in in vivo cell therapies; the FDA piloting a faster route into first-in-human trials, and other key signals.

After a decade of AI drug discovery, two of its vocal voices assess progress

Perhaps unusually for August, there is a lot of business and research activity in the artificial intelligence (AI) drug discovery space. Notably, within four days of each other, two respectable voices in the AI drug discovery space independently published two important pieces about the state of progress, where industry stands about all sorts of AI tools as of 2026, and where it may take us from here.

On August 3, insitro founder and celebrity in the field of artificial intelligence published “Drug Discovery Has No Magic Wands.” She splits development broadly into three stages:

… and argues AI effort has concentrated overwhelmingly on the middle one, molecular design, while more than 90% of clinical failures come from picking the wrong mechanism. Even total success on the AI-addressable portion, she notes, touches only about half a development timeline; the cell atlases needed to attack stage one are “orders of magnitude too small.”

On August 7, Dr. Andreas Bender, Professor for Machine Learning in Medicine at Khalifa University, and a large co-author group published “Artificial intelligence in drug discovery — what it is, where we stand and the path forward” in Nature Reviews Drug Discovery.

Its verdict on a decade of work: clinically relevant impact has been “disappointingly limited,” caused by too little focus on clinical translation, the context-dependent nature of real biological data, and problem definitions too vague for the decisions the models are meant to inform. The prescription is a shift from “technology push” to “science pull,” and benchmarks that measure improved decision-making rather than model performance.

Neither piece cites the other, the convergence seems to be independent, which likely signals a growing industry consensus about the current state of AI in drug research.

⚡ This issue is brought to you by SYDRA

The Brief

💰 Follow the Money

🔹Dash Bio raised a $30M Series A led by Oak HC/FT for its robotics-run bioanalysis lab — preclinical assays back in days, not months.

🔹Pathos AI in-licensed two oncology assets in one day for a reported ~$2.2B — a TROP2/HER3 bispecific ADC from Alphamab and an ERα PROTAC from AstraZeneca.

🔹Bruker took a majority stake in organ-on-a-chip maker MIMETAS (terms undisclosed). Why it matters: positions Bruker in the New Approach Methodologies push to replace animal testing.

🔹Anthropic opened rare disease research grants — up to $50,000 in Claude credits over six months, two tracks, applications closed August 2.

🤝 Deals & Alliances

🔹Bristol Myers Squibb is deploying Schrödinger’s agentic AI co-scientist Bunsen across its research organization, plus the RetroSynth synthesis planner. Terms undisclosed.

🔹Genentech exercised its first option on a Recursion-discovered neuroscience target, moving it into joint small-molecule discovery. The numbers: first target out of a collaboration that has paid Recursion $216M to date, with up to 40 possible programs. — link

🔹ICON, a top-five CRO, signed a multi-year deal to embed Claude across the clinical trial lifecycle. Backstory: caps a year of Anthropic life-science moves — Claude for Life Sciences, the ~$400M Coefficient Bio buy, Claude Science, and hiring AlphaFold’s John Jumper.

🔹J&J paid $785M for an option to acquire in vivo CAR-T player Sail Biomedicines for a further $2.58B. Backstory: the fifth major in vivo CAR-T deal since Lilly’s $7B Kelonia buy in April thinned the field of outright targets.

🔹Merck KGaA adopted Evinova‘s AI-native clinical development platform — AstraZeneca’s in-house health-tech business selling to a competitor.

🔹Pharma is pooling its structural data. C&EN on the Federated OpenFold3 Initiative, where AbbVie, J&J, BMS, Takeda and Astex train a shared open model without exposing IP.

🔬 Science to Watch

🔹A CRISPR enzyme that kills cancer cells by shredding their own DNA. Doudna’s group published Cas12a2 in Nature: triggered by a mutant RNA transcript, it destroys the host cell’s chromatin, killing p53-mutant cells while sparing healthy ones. Preclinical.

🔹An AI biologist called XunZi flagged CHK2 as a Parkinson’s target; inhibiting it rescued dopaminergic neuron loss and motor deficits in mice. Nature Biomedical Engineering, Aug 4.

🔹Open-weights models that read chemistry out of patents. Edison Scientific’s MarkushGlyph parses Markush structures at 58–64% accuracy; single-structure recognition hits 93.8%. What it is: Markush diagrams are how one patent claims millions of compounds.

🔹Clinical AI’s dangerous failure is omission, not fabrication. The NOHARM benchmark found errors of omission account for >80% of severe errors across 20 LLMs and 4 RAG clinical tools. Preprint.

🔹Recursion says protocol simulation against real patient data expanded trial eligibility 10–40%, with enrollment 30–60% above historical projections. Vendor-reported.

🔹Insilico launched a benchmark for AI drug discovery models — 300+ evaluations across two suites, live since July 30. No leaderboard scores published yet.

🔹~50 authors proposed an L0–L5 autonomy ladder for laboratories, arguing self-driving labs need a shared “world model” layer. Leskovec, Uhler, Ioannidis and Shah among the signatories.

📋 From the Regulators

🔹FDA is piloting a faster route into first-in-human trials. The Expedited IND Pilot would let sponsors use pre-vetted research institutions and file INDs on a rolling basis; the agency estimates clarified Phase 1 CMC expectations alone could save 6–12 months.

🔹A companion draft guidance shifts first-in-human dose selection toward QSP modelling and away from animal toxicology.

Leave a comment

⚡In partnership with SYDRA

Most AI drug-discovery companies begin with a target and use algorithms to optimize molecules around an assumption. SYDRA reverses the logic.

Our proprietary AI selects and generates novel chemistry; whole-organism lifespan assays reveal which molecules actually affect aging biology; and only the phenotype-proven winners advance into aged-mouse studies, human-cell target deconvolution and disease positioning.

Starting from six million molecules, we identified five significant lifespan hits, advanced two lead programs into late-life mouse pilots, and created a new generative chemistry cohort now entering synthesis. We don’t use AI to decorate a target hypothesis. We let living biology decide which AI discoveries deserve to become medicines.

SYDRA is now raising a CHF 2.6 million pre-seed round to accelerate lead programs toward pharma partnering.

A Thing To Know This Week

In vivo CAR-T, and why everyone is buying it

Conventional CAR-T extracts a patient’s T cells, engineers them in a facility to recognise a cancer target, and infuses them back. It works, at several hundred thousand dollars per patient, over weeks, at specialist centres only.

In contrast, in vivo CAR-T does the engineering inside the body: inject a lipid nanoparticle or targeted viral vector that reprograms T cells where they already live. No extraction and no manufacturing slot, a body becomes, basically, its own bioreactor for the therapy. If it works, a cell therapy becomes almost as simple as a regular injection.

That “if” is what eighteen months of dealmaking has been pricing on the level of strategic M&As.

AstraZeneca, AbbVie and Gilead each bought a platform in 2025; Lilly bought two in early 2026, for up to $7B in total. Most of these assets are preclinical — pharma is arguably buying delivery technology on the bet that the biology is proven and only the logistics are mostly unsolved.

Interestingly, Johnson & Johnson appears to be more cautious. J&J’s $3.5 billion deal with Sail Biomedicines serves as the textbook definition of an "option-first" structural hedge. Rather than acquiring an in vivo company outright at a massive premium like others, J&J structured a massive option agreement that secures early platform access while shifting early clinical risk entirely onto the partner.

Share

Strategic Signals…

🔗 DeepMind dismantled the AlphaFold team — and it matters less than the headline suggests.

🔗 Ramy Farid called AI in drug discovery “nonsense” in 2024. In July he shipped an AI co-scientist and said he couldn’t be more pleased to be wrong about AI in the past.

🔗 “Forward deployed scientist” is becoming a real job title — Chai, Phylo, Lila and Periodic Labs are all hiring it.

🔗 China's share of first drug launches rose from 3.3% to 28.7% between 2004 and 2024 while Europe's fell from 26.7% to 8.5%, per a new analysis. US is still leading the pack but declined from 45.9% to 37.8%.

That was last week, and have a great one ahead!

Forward this to a colleague who should be tracking this space,

Cheers!

— Andrii

Share

Read also:

Could Anthropic Disrupt the Techbio Companies Built for Pharma/Biotech? (Also: The Numbers Behind China’s Biopharma Ascent)

Is the Future of AI Drug Discovery Hybrid?

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

Cancer as a Data Problem: What AI Is Doing in Oncology

But despite the obvious impressiveness and importance of this move, the part that Anthropic is addressing is the one that was never the most expensive/complex part of the drug development business, in my opinion.

Designing a promising molecule has been getting cheaper for a decade, as well as target selection. That is not where drug discovery programs fail, generally speaking. Drug programs fail in toxicity and unexpected systemic effects on body, lack of efficacy, in manufacturing, in failing to enroll enough/right patients, in trials that fail because the biology was misjudged or the population was selected in a sloppy trial design (lack of biomarkers, etc.).

As far as I can tell, Anthropic has limited clinical apparatus, no manufacturing, no regulatory track record, and, likely, no answer yet to “what do you do if you find something that works?” Yes, sure, they can design drugs really fast, but then what? How does their model differentiate against later stages, where the role of AI is surprisingly small at the moment?

Claude Science can probably move “target to shortlist in minutes,” and so it compresses the cheapest, fastest, most-solved stage of the pipeline. It is certainly a milestone, certainly impressive, but have they solved drug discovery with it? Probably not yet.

Now, what does it mean for the AI-native biotechs, the likes of Iambic, Recursion, Insilico Medicine, OWKIN, Isomorphic Labs, Cradle, NOETIK, Xaira, SandboxAQ, etc.)?

Well, their moat was never the model alone, I think. It is this:

This article is for paid subscribers of BiopharmaTrend Pro membership.

Keep Reading -->

I spent last week in Amsterdam, attending HLTH Europe 2026, arguably Europe’s flagship healthcare innovation event, hosting more than 5,000 attendees. It is my second HLTH experience, and two additions to this year’s program stood out: last year’s Pharma & Life Sciences Spotlight grew into a full two-day Global Pharma Summit aimed at C-suite from 30-plus pharma and biotech organizations across 20-plus countries, and the floor gained a dedicated AI @ HLTH zone with its own stage dedicated completely to the progress of artificial intelligence in healthcare and clinical research.

The organizers framed the agenda around whether healthcare AI has broken the Gartner hype cycle, skipping the trough of disillusionment and landing straight into clinical workflows, patient portals, and everyday tools, or not…? The reality is, as always, nuanced and ambiguous, but whatever it is, I have gathered several observations from various interviews and discussions, the HLTH Europe show floor, and the announcements timed to the event, that might be shaping this area in the second half of 2026 and beyond.

Keep Reading -->

Everyone is talking about frontier AI models and agents. But the most interesting conversations I had in Barcelona earlier this month during two cutting-edge bioinformatics events pointed in a different direction.

I attended two events back-to-back: the Alchemistry Workshop on Free Energy Methods (May 4–6) and the inaugural CoFold Summit (May 6).

The first is the established annual conference for physics-based drug design, with participation and sponsorships from companies like Schrödinger, AstraZeneca, Cresset, OpenBioSim, etc. The second brought together the teams building deep learning co-folding models — Isomorphic Labs, Boltz, OpenFold, RoseTTAFold, SandboxAQ, and others. Same city, same week, overlapping audiences.

Frontier AI models alone won’t get you far in biology. We need to invest in physics-grounded tools and methods.

Anyway, speaking about free energy perturbation methods, they have become the industrial workhorse for binding affinity prediction, and GPU acceleration has moved them from supercomputers to everyday pharma workflows. But FEP needs good starting structures, and it struggles with structurally diverse compounds coming out of generative AI pipelines. On the co-folding side, models like Boltz-1x are making real progress on the chemical validity of predicted poses. But they still default to well-represented binding sites from training data and can’t reliably score what they generate. Allosteric pockets, novel targets, anything underrepresented, still a major challenge.

The pattern kept coming up in different sessions and hallway conversations.

This article is for paid subscribers of BiopharmaTrend Pro membership.

Keep Reading -->

Their thesis is a bit different from what most AI biopharma companies are doing. Instead of designing new molecules, Biossil uses AI to dig through late-stage clinical failures and figure out which patient subgroups those drugs should have actually been tested on. Ten molecules were acquired while in stealth mode over three years. Trials running in everything from glioblastoma to Alzheimer’s.

That’s not drug repurposing in the classic sense — taking an approved drug and finding it a new indication, like thalidomide going from its original (disastrous) use to multiple myeloma, or metformin being studied in cancer.

Biossil is doing something more subtle: same molecule, same disease, just a more precisely defined subset of patients. The argument is that many drugs “failed” trials only in the “aggregate”, averaged across a heterogeneous population where a real signal got buried.

And they’re not alone. A cluster of companies, each with different technical approaches and varying levels of clinical evidence, is converging on a shared conviction: the pharma industry’s 90%+ clinical failure rate isn’t just a scientific problem. It’s partly an analytical one. The tools to find the right patients simply weren’t good enough, until now.

This is a piece about that convergence. We’ll map who’s doing what, how the approaches differ, what’s actually been validated, and whether the thesis holds up under scrutiny.

In this issue: The Logic of Drug Rescue — The Landscape: Who’s Doing What — A Closer Look at the Frontrunners — The Roivant Precedent — What Doesn’t Work (Yet) — Looking Ahead

This article is for paid subscribers of BiopharmaTrend Pro membership.

Keep Reading -->

Recent developments suggest that the longevity industry is beginning to outgrow its infancy and move into a more mature formative stage. What started with academic aging research, biohacking, mouse studies, and occasional n=1 self-experiments is increasingly evolving into a well-funded biotech ecosystem focused on translating aging-related mechanisms into clinical development. A prominent example is Life BioSciences, the cell-rejuvenation company co-founded by aging researcher David Sinclair, which recently raised an $80 million Series D round to advance its first-in-human clinical program in optic neuropathies. Beyond cellular rejuvenation, a growing number of biotechs are developing therapies aimed at other hallmarks of aging, from senescence to metabolic dysregulation.

At the same time, despite these encouraging signs of momentum, the longevity industry still faces fundamental challenges, including the absence of a unified hypothesis of aging and the lack of robust proxy markers that can reliably measure biological aging and therapeutic impact.

The promise and challenges of the budding longevity industry

For most of human history, living longer mainly meant not dying early. Over the past two centuries, particularly in the 20th century, clean water and sanitation, improved nutrition, vaccines and antibiotics, and safer maternal, neonatal, and emergency care helped push average life expectancy up by decades.

In many high-income settings, however, the pace of improvement has slowed since the 1990s, while at the same time the burden of chronic disease has increased significantly. This raises a new question: how can we live longer, individually and as a society, without simply shifting the burden into more years lived with chronic disease?

Researchers are increasingly reframing the problem away from purely disease-by-disease fixes and toward the underlying biology of aging that contributes to multiple conditions, such as cancer, cardiovascular, neurodegenerative, and metabolic disease. While the field is only beginning to untangle the complex, and likely multifactorial process of aging, a fledgling longevity industry has emerged around various hypotheses of aging mechanisms. This longevity stack spans different layers from biomarkers and aging clocks that aim to measure biological aging, to therapeutics that target aging-linked mechanisms, as well as care and delivery models (clinics, digital platforms, consulting) that package measurement and interventions into services (Figure 1).

This article is for paid subscribers of BiopharmaTrend Pro membership.

Keep Reading -->

Also, if you're at Drug Discovery Chemistry in San Diego this week, our board advisor Andrew Lozoniuk is there and happy to chat!

Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

Subscribe now

🤖 AI x Bio

(AI applications in drug discovery, biotech, and healthcare)

Agents at AACR 2026

At the American Association for Cancer Research annual meeting this year, out of ~9,600 accepted abstracts roughly 1,100 (1 in 8) are AI-related, according to Miraei AI.

Within that: ~160 reference LLMs, generative AI, or agentic systems; ~140 involve multimodal approaches; ~225 cite clinical validation or deployment language. The Miraei team’s read is that AI in oncology is shifting from isolated proof-of-concept models to embedded operational systems.

The leaderboard of AI-related abstract affiliations at AACR 2026, split by academic institutions and commercial entities:

AACR will host a dedicated “Agentic AI in Cancer” session: MD Anderson will debut “Charles,” a self-critical agentic drug discovery analyst built around traceability and hallucination mitigation. MSK will show an MCP-enabled AI agent in its genomics platform automating multimodal analysis at petabyte scale. AstraZeneca (the most-represented biopharma in the AI corpus with ~10 abstracts) will present an agentic system for RNA-seq pipeline automation. (AstraZeneca also acquired Modella AI earlier this year, whose Judith agent automates pathology image analysis workflows.)

Foundation models were still a distinct thread. Bioptimus delivered three abstracts including two late-breakers—H-optimus-1 for computational histopathology and M-Optimus-1, a multimodal model unifying H&E imaging, bulk and single-cell RNA-seq, and spatial transcriptomics. A UCSD–Lunit collaboration will present a foundation model of cancer genotype predicting therapeutic response.

Around 225 abstracts use language suggesting clinical validation or deployment, not just modeling accuracy on retrospective data. A separate “LLMs in the Clinic” session (~30 abstracts) will focus on infrastructure problems: trial matching, EHR extraction, pathology report abstraction, patient message triage. A University of New Mexico study reported cancer patients preferred ChatGPT-generated responses to physician-authored ones in a blinded comparison.

This tracks with what we covered in our February deep dive on AI in oncology. The pattern across ~1,100 abstracts is the same shift we described there: from standalone models to infrastructure. Agentic systems managing multi-step workflows, LLMs doing clinical plumbing, foundation models trained at institutional scale. Commercial actors (Tempus, Labcorp, Lunit, Natera, BMS) are showing operational work; the startup layer (Bioptimus, Data4Cure, Keiji AI, EXoPERT) is landing late-breaking slots.

AI in the pipeline

A few recent moves worth noting across clinical trials, pharma infrastructure, and AI-designed therapeutics.

🔹 Massive Bio published a prospective study in ESMO Real World Data and Digital Oncology with 3,804 cancer patients, 17,000+ oncologist-confirmed trial matches, 4x faster matching (120-to-30 min per patient), outperforming zero-shot GPT-4 and GPT-4o.

🔹 CRO Fortrea released an AI suite for clinical workflow automation, part of CEO Anshul Thakral’s turnaround strategy since taking over in June 2025.

🔹 Tempus expanded its collaboration with Gilead with enterprise-wide access to Tempus’s Lens platform for oncology R&D, continuing the pattern of large pharma outsourcing data infrastructure to specialized AI/RWE companies.

🔹 Astellas exercised its option to license a Dyno Therapeutics AAV capsid for skeletal muscle gene delivery, following Roche’s CNS capsid license in January 2025 — making Dyno the first company to license AI-designed capsids for both CNS and muscle ($15M fee plus milestones and royalties).

💰 Money Flows

(Funding rounds, IPOs, and M&A for startups and smaller companies)

Daniel Strasser, Executive Director and Therapeutic Area Biomarker Head at Novartis, posted a tracking that puts the pace of AI-pharma deals in perspective.

• Between 2021 and 2024, top pharma signed 2–4 AI deals per year. In 2025: 15. At Q1 2026 pace, the year is tracking toward ~24 — a 12x increase in five years.

• Combined potential deal value across 34 tracked partnerships exceeds $78B. All top-10 pharma like Lilly, Novartis, Roche, Sanofi, AstraZeneca, J&J, Novo Nordisk, Takeda are in with billion-dollar-scale commitments.

• NVIDIA alone ($4.4T) is worth more than all top-10 pharma combined ($3.5T), and the AI-pharma pure-plays signing these deals collectively sit around $23B.

Strasser's question is—which side captures the value?

Cell reprogramming in the clinic

Longevity raise—Life Biosciences (co-founded by David Sinclair) closed an $80M Series D for the Phase 1 trial of ER-100—reportedly the first partial epigenetic reprogramming therapy to receive FDA IND clearance. ER-100 uses AAV2 vectors encoding three Yamanaka factors and an oral small-molecule switch, targeting retinal ganglion cells in optic neuropathies.

Q1 2026 was strong for longevity biotech overall ($3.7B across 49 deals, 56% YoY). Competitors remain largely preclinical:

• Altos Labs (~$3B, stealthy)

• Retro Biosciences ($1B+ Series A, clinic-stage autophagy)

• NewLimit ($130M Series B, liver)

Life Bio’s differentiator is clinical specificity, though the sector still has to prove it can produce credible aging biomarkers alongside disease-specific efficacy.

Europe’s largest independent biopharma raise?

Jeito Capital closed Fund II at over €1B (~$1.2B). The fund targets 15-to-20 clinical-stage investments at up to €150M each, addressing a funding tier where European options have historically been thin.Fund I (€534M, 2021) exits include Eyebiotech (Merck, $1.3B) and Hi-Bio (Biogen, $1.15B).

For context: between 2015 and mid-2025, EU biotech startups attracted €25B in venture capital versus €219B in the US, and 66 of the 67 EU companies that went public over the past six years listed on foreign exchanges.

🔹 A digital therapeutic for negative symptoms of schizophrenia—Click Therapeutics got $50M Series D from Boehringer Ingelheim plus full commercial rights for CT-155. Phase III CONVOKE met its primary endpoint (p=0.0003, Cohen’s d = –0.36). Click already has two authorized digital therapeutics (migraine, major depressive disorder) and 63 patents.

🔹 The OpenAI Foundation announced $100M+ in initial grants for Alzheimer's research across five tracks: causal mapping (Arc Institute), AI-assisted drug design (Institute for Protein Design + Mass General Brigham), open datasets (EvE Bio), biomarkers (UCSF), and off-patent treatment testing (Harvard). David Baker's IPD team says they've already engineered molecules that engage and degrade Alzheimer's-relevant targets using AI protein design models. The off-patent track is notable: lithium orotate and the shingles vaccine both have suggestive evidence but no private-sector incentive to fund proper trials. More grants expected through 2026.

🏛️ Bioeconomy & Society

(News on centers, regulatory updates, and broader biotech ecosystem developments)

🔹 Neuropacs received De Novo classification from the FDA for a diffusion MRI–based AI diagnostic aid that helps differentiate atypical Parkinsonian syndromes (MSAp, PSP) from Parkinson’s disease. The classification establishes a new regulatory category “Parkinsonian syndrome diagnostic aid,” and is the first device in it. The tool draws on 15+ years of research and a prospective multicenter study across 21 centers (Parkinson Study Group, published in JAMA Neurology). Neuropacs recently closed a $1M seed round.

⚙️ Other Tech

(Innovations across quantum computing, BCIs, gene editing, and more)

🔹 Can it deliver? Quantum computing is drawing pharma bets. Amgen has worked with Quantinuum since ~2018 and invested in its $300M round in 2023—building competence now for utility that’s 5–10 years out. Novo Holdings is assembling a portfolio (Sparrow Quantum, Phasecraft) and says it’s “confident tangible quantum advantage for defined applications will be achieved before 2030.” VC in quantum hit $3.8B in 2025. Merck KGaA’s Thomas Ehmer offered the counterpoint: the field is “always three years away,” and whether the killer app justifies advance competence-building remains open.

🔹 Miltenyi Biomedicine’s CAR-T drove three autoimmune diseases into simultaneous remission in a single patient—AIHA, antiphospholipid syndrome, and immune thrombocytopenia, all B-cell–driven and previously treatment-resistant. Eleven months out remission holds. Kyverna Therapeutics is likely first to an autoimmune CAR-T approval (stiff-person syndrome, FDA submission expected H1 2026).

Subscribe now

Cover: Antique celestial map illustration with mythological constellations and star charts from an early 19th century astronomy book, published in London in 1822. iStock

Read also:

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

Cancer as a Data Problem: What AI Is Doing in Oncology

Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

Subscribe now

🤖 AI x Bio

(AI applications in drug discovery, biotech, and healthcare)

The CRO intelligence layer remains unclaimed

Reuters reported that CRO stocks (IQVIA, Medpace, Charles River) have dropped sharply since Anthropic launched AI agents in February. The authors’ conclusion is that the market is overreacting, AI can’t replace patient recruitment and site execution, CROs will be broadly fine.

The near-term read is probably right, but the longer-term question is more interesting: what happens when the design layer of trials (protocol optimization, site selection, patient matching) gets commoditized by AI tools that sponsors can run themselves?

As noted recently, strategic CROs sell execution and intelligence bundled together. Unbundle that, and they’re competing on operations and logistics alone, which is a different kind of business that’s priced differently.

There’s also a second angle: if AI lowers the intelligence barrier, a new class of leaner, tech-native CROs could assemble trial infrastructure on demand, more platform than services firm. This structural shift is explored in our 2023 BiopharmaTrend analysis of the CRO industry, though agentic capabilities at the time were still early. Nobody is cutting CRO spend yet. But “no evidence of cuts” is not the same as “no shift in leverage.”

Agentic AI in drug discovery is already delivering value, but not where the hype points

📝 A new review in Drug Discovery Today covers eight case studies from companies including Coincidence Labs, Human Chemical, Potato, Kiin Bio, Augmented Nature, onepot, Plex Research, and Convexia. Some reported specific quantitative benchmarks, not just “we used AI and it was faster.”

The most convincing gains are in the tedious middle: literature synthesis, protocol generation, assay design. Potato’s Tater agent took the design cycle for a qPCR assay from literature review through to executable lab automation code from 1-4 months to under two hours. Empirical validation was still needed, but the bottleneck it removes is the coordination tax between “we know what assay we need” and “here’s a draft protocol ready for the bench.”

Note: there’s missing cost data: every case study reports time savings and none report what the agentic infrastructure costs to build and maintain.

The authors flag that current benchmarks capture whether an agent got the right answer but not whether the reasoning behind it was sound. And a separate evaluation from Princeton and Cornell suggests the problem runs deeper: across 14 AI agent systems, agents that could solve a task often failed on repeated attempts under identical conditions. Bigger models didn’t reliably fix this.

Our read: the real value right now is compressing the overhead between steps that already work on their own, not autonomous science (yet). We went deeper on where agents are actually being deployed, what’s breaking, and what alternatives look like in our recent deep dive:

➕ Fay Lin at GEN also surveyed agentic platforms now on the life sciences market after NVIDIA GTC: Kosmos, LabOS, Latent-Y, Dyno Psi-Phi, and others.

Share

Is AI replacing radiologists?

The “AI replacing radiologists” discourse resurfaced this week after NYC Health + Hospitals CEO Mitchell Katz said AI could replace a significant portion of radiology work—initial AI reads, radiologist review of flagged abnormals. The Stanford-Harvard NOHARM study showed top LLMs outperforming generalist physicians on clinical safety, and the numbers look good in controlled settings.

Bo Wang, Chief AI Scientist at University Health Network (Canada’s largest research hospital network), offered a corrective from deployment experience:

The accuracy is real in narrow tasks. But NOHARM found 76.6% of AI errors were omissions = things the model never flagged. In a hospital, a missed finding propagates: downstream physicians trust the AI read, the patient waits, the window closes. (One caveat: NOHARM evaluated LLMs on clinical recommendations across ten specialties, not imaging-specific AI, though the omission pattern maps to radiology concerns.)

No hospital system has solved the accountability question either. When an AI-assisted diagnosis causes harm, the liability chain between physician, hospital, and vendor has no established framework. Which is not regulatory delay but “...a fundamental gap in the infrastructure for AI-in-medicine.”

In Wang’s view, here’s what’s actually happening: AI is reshaping the radiology job, not eliminating it. Routine reads get faster, time shifts to complex cases and clinical correlation. The capability question is nearly answered, while the deployment question has barely been asked.

AI clinical operations partnerships keep stacking

🔹 Bristol Myers Squibb and Faro Health announced a multi-year deal to apply AI across clinical trial design and protocol workflows. Faro structures protocols into standardized, machine-readable formats and benchmarks them against datasets like Tufts CSDD, with work spanning both early- and late-phase programs. Faro is not an isolated instance—this year BMS also partnered with Evinova on trial design and with Microsoft on AI-based patient identification.

🔹Tempus reports that its ALERT trial, conducted with Medtronic across 35 U.S. hospitals, showed that AI-driven alerts embedded in EHRs improved detection and treatment of severe heart valve diseases. The study had 765 clinicians, over 2,000 echocardiograms, and used Tempus’s “Next” to analyze echo reports in real time with NLP and notify clinicians about high-risk findings.

🔹 OpenEvidence partnered with Tandem to connect real-time, AI-driven clinical decision support with automated prescription and insurance workflows. OpenEvidence functions as a clinical search layer, synthesizing evidence from sources like NEJM, JAMA, and Cochrane to support treatment decisions at the point of care, with over 1 million daily consultations in the U.S. Described as “ChatGPT for doctors,” OE raised $250M at a $12B valuation this year and generates over $100M in annual revenue, in part through ads embedded in the platform.

AI in the pipeline layer

🔹 Boehringer Ingelheim is expanding its collaboration with OpenProtein.AI to push AI-driven antibody discovery further into its development workflows. The pairing of a large pharma with deep therapeutic pipelines and a startup built around protein foundation models is becoming a common pattern.

🔹 Infinimmune and Merck entered a discovery and development partnership—Infinimmune will use its platform to screen human immune cells and identify naturally occurring antibodies, then refine them with AI tools. Merck gains exclusive commercialization rights, with potential milestone payments up to $838M across multiple undisclosed disease targets.

🔹 Evotec, Germany-based drug R&D company operating a large network of partnered R&D platforms, appointed Ashiq Khan as chief commercial officer, bringing AI and robotics experience from previous roles at Iktos and Schrödinger. The hire is part of Evotec’s broader transformation plan announced last month, which named commercial execution as a core focus.

🔹 Insilico Medicine partner TaiGen Biotechnology achieved first-subject enrollment and dosing in the Phase 1 trial of ISM4808—an oral HIF-PHD inhibitor for CKD-related anemia discovered with Insilico’s generative AI platform for small-molecule design in under three months (from the December 2025 in-licensing). Another data point for Insilico’s growing list of partnered programs moving into clinic.

🔹 Enveda Biosciences reported early clinical data from its Phase 1b trial of an oral candidate discovered via the company’s AI platform, in moderate-to-severe atopic dermatitis. Nine patients treated at 800 mg daily for 28 days showed rapid reductions in disease severity and itch within the first week, with improvements continuing after treatment cessation. No serious adverse events were reported. ENV-294 is a non-degrading stabilizer (LOCKTAC) showing biomarker activity across Th1, Th2, and Th17 pathways—broader modulation than single-cytokine-axis therapies. Enveda plans Phase 2a trials in atopic dermatitis and asthma, with Phase 2b targeted for mid-2026.

🔹 Lantern Pharma, the AI-driven oncology company known for its RADR AI platform, is joining the agentic wave with withZeta.ai—a multi-agentic co-scientist platform for rare cancer drug discovery that autonomously queries proprietary knowledge bases, clinical trial databases, and molecular libraries. The company will demo the platform at an investor briefing on April 9.

💰 Money Flows

(Funding rounds, IPOs, and M&A for startups and smaller companies)

Pharma went on a $25.5B, eight-day acquisition spree. Six pharmaceutical companies signed acquisitions worth up to $25.5B by the start of this month, including two $5B+ upfront deals disclosed on the same Tuesday morning: Eli Lilly’s $6.3B bet on Centessa Pharmaceuticals’ orexin drug candidates and Biogen’s acquisition of Apellis Pharmaceuticals. For context, there were a total of seven deals of that magnitude in all of 2025.

Across March, biopharmas lined up 10 acquisitions worth up to $31.5B, per Endpoints News. Eight included contingent value rights or milestone payments—a dealmaking structure that has become more common in recent years. Year-to-date, Endpoints has tracked up to $48B in total potential deal value across 19 biopharma acquisitions. Société Générale tallied $126B in total deal value for all of 2025, well above the 10-year average of $85B. If the current pace holds, 2026 could surpass it.

🔹 Blackstone has collected $6.3B for its latest life sciences fund, which has already committed roughly $2B over the past year for an ADC at Merck, an anti-TL1A at Teva-partnered Sanofi, and other deals. The fund invests in approximately 3% of deals it evaluates, targeting therapies with $1B+ peak sales potential.

🔹 Anthropic acquired Coefficient Bio, a stealth biotech AI startup, for $400M in stock. Coefficient Bio was founded eight months ago by Samuel Stanton and Nathan C. Frey, both from Genentech’s Prescient Design group. The team of roughly 10 people was using AI for drug discovery planning, clinical regulatory strategy, candidate identification, and biomolecule modeling. As Jonah Probell of Lexi Ventures noted, the deal generated a reported 38,513% paper IRR for Dimension, the VC firm that backed them. Anthropic’s head of Healthcare & Life Sciences Eric Kauderer-Abrams stated the goal rather plainly: Anthropic wants a meaningful share of all life science work globally to run on Claude. The team is expected to join Anthropic’s health and life sciences unit.

🔹 Generare, a Paris-based techbio building a discovery platform around microbial DNA, raised €20M in a Series A co-led by Alven and daphni. The company identifies biosynthetic gene clusters in microbial genomes, expresses them in lab systems, and characterizes the resulting molecules, reporting over 200 previously uncharacterized small molecules per cycle versus an industry baseline of ~45. Most screening libraries draw from an estimated 3% of accessible chemical space, and Generare is betting that new chemistry from unexplored microbial genomes, not just more data from known chemistry, is what AI-driven drug discovery needs. The team of ~25 aims to scale output to 2,000+ molecules by 2027. Founded in 2023 by Guillaume Vandenesch and Vincent Libis.

🔹 Scala Biodesign, an Israeli computational protein design platform, raised $16M in a Series A led by Grove Ventures, joined by TLV Partners, Deep Insight, and the Israel Innovation Authority. The platform combines LLMs, physics-based modeling, and evolutionary data to help clients design and optimize proteins for biologic, enzyme, antibody, and vaccine therapies. It aims to become routine infrastructure across the full development pipeline rather than a one-off discovery tool.

🏛️ Bioeconomy & Society

(News on centers, regulatory updates, and broader biotech ecosystem developments)

The EMA just issued its first draft qualification opinion for virtual control groups as a New Approach Methodology.

Context is rather narrow: rat dose-range finding studies, non-GLP, replacing concurrent control groups with virtual comparators built from historical data. It’s a narrow starting point, but it sets a regulatory blueprint for every study design where live controls exist by convention rather than scientific necessity.

Stefano Gaburro, who leads the Pistoia Alliance Minimal Metadata Set Working Group, noted that the real bottleneck was never the algorithm, but the metadata. Inconsistent annotation across institutions prevented historical control data from being pooled and reused. Pistoia’s MNMS framework, notes Stefano, was built specifically to create the data infrastructure that virtual control groups require. A cross-pharma proof of concept with five companies, using 24/7 home cage monitoring data, is ongoing. The EMA consultation runs until May 12, 2026.

Around the same time, Joseph Wu and colleagues published a review in Cell on NAMs for drug discovery, framing the evolution from 2D stem cell systems through 3D organoid models to AI-driven in silico platforms. The review argues that the next 30 years of drug development will see animal models transition from a central role to a supporting one, following the 3Rs principle, replaced increasingly by human-centric NAMs powered by multi-omics data and AI pipelines. Between the EMA opinion and this review, the regulatory and scientific architecture for reducing animal use in drug development is being built in parallel.

⚙️ Other Tech

(Innovations across quantum computing, BCIs, gene editing, and more)

🔹 Bone marrow by the moon—NASA’s Artemis II mission is carrying AVATAR (A Virtual Astronaut Tissue Analog Response)—organ-on-chip devices containing bone marrow cells from crew members to study how deep space radiation and microgravity affect human health during the approximately 10-day lunar journey.

Bone marrow was selected for its radiosensitivity and centrality to immune function. The battery-powered payload, developed by Space Tango, will maintain automated environmental controls during flight. Upon return, Emulate researchers will use single-cell RNA sequencing to compare flight samples against ground-based controls. NASA has extended organ-on-chip viability to six months minimum to enable observation of long-term disease and treatment responses.

Share

🚀 A New Kid on the Dura

Epia Neuro launched from San Francisco this week with a minimally invasive brain-computer interface designed to help stroke survivors regain motor function by reading what the brain intends to do and translating that into movement.

The founder is Michel Maharbiz, who previously built iota Biosciences, a bioelectronics startup later acquired by Astellas. The team spans engineering and neuroscience, with backgrounds in both research and clinical system development.

Stroke is the entry point: roughly 690,000 cases annually in the U.S., with around 60,000 patients potentially eligible for device-based intervention. Epia plans first-in-human demonstrations with the Department of Neurosurgery at Lenox Hill Hospital.

The technical differentiator is a read/write architecture where the device can both record brain signals and deliver stimulation, spanning cortical and deeper brain regions. This contrasts with earlier generations of primarily read-only BCIs. The implant is placed within the skull without penetrating brain tissue via a roughly one-hour surgery. Onboard algorithms adapt to each patient’s activity patterns over time, with wireless recharging and upgradeable architecture built in for long-term use.

Early on, the device supports rehabilitation, reinforcing intended movements through feedback and stimulation while recovery is still possible. As recovery plateaus, it shifts into an assistive role, driving wearable external devices for tasks like gripping objects. The company is also signaling longer-term ambitions in cognitive decline, Parkinson’s, and neuropsychiatric conditions.

The BCI landscape we’re tracking is moving fast. Just last month we covered how Neuracle’s minimally invasive device became the first commercially approved BCI in China, followed by increased funding activity:

• StairMed raised $73M for its robot-inserted flexible electrode system;

• Gestala raised $21M for a non-invasive ultrasound-based platform.

• In the U.S., Merge Labs (backed by Sam Altman) raised $252M for ultrasound-based BCIs

• Nia Therapeutics received FDA breakthrough device designation for an AI-guided memory loss BCI

• Neuralink, the largest actor, plans high-volume production and a near-fully automated implantation workflow in 2026 following a $650M Series E.

We covered the broader neurotech landscape in our Neurotech Review here.

Subscribe now

Read also:

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

Cancer as a Data Problem: What AI Is Doing in Oncology

A team at Stanford recently posted a preprint describing a system they called “Virtual Biotech”: a coordinated squad of AI agents organized to mirror a real drug discovery company, complete with a virtual Chief Scientific Officer, specialized scientist agents, and over 100 tools for querying biomedical databases.

For their headline demonstration, they deployed over 37,000 agents in parallel, each one tasked with annotating a single clinical trial, linking therapeutic targets to genomic and single-cell transcriptomic features. The resulting dataset spans 55,984 trials. The analysis turned up what the authors call previously unreported associations: drugs targeting cell-type-specific genes were 40% more likely to advance from Phase I to Phase II, 48% more likely to reach market, and showed 32% lower adverse event rates.

Virtual Biotech workflow. User query > CSO clarification and briefing preparation > specialized scientist agents > scientific reviewer > synthesis or revision. Source: Zhang et al., bioRxiv, Feb. 23, 2026.

In another case study, the system pulled together genetics, transcriptomics, and clinical data on B7-H3 in lung cancer and landed on an antibody-drug conjugate strategy, the same bet several pharma companies are already running in the clinic. It also flagged liabilities and differentiation angles. The whole thing reportedly cost $46 in API credits and took less than a day.

The Virtual Biotech is one lab’s preprint, but it lands in what we recently likened to a “gold rush”—agents are being deployed across clinical operations, translational biology, antibody design, and regulatory workflows. Major pharma companies are in an apparent compute arms race, stacking GPU clusters and billion-dollar AI partnerships within months of each other. Startups backed by hundreds of millions are launching agent-focused platforms. NVIDIA’s Jensen Huang even went so far as to declare agentic AI “the new computer” at this year’s GTC.

Whether the implementations match is another question. In a recent experiment, researcher Liang Chang asked — ”Can AI make better decisions than pharma executives?” and sent AI agent teams back to a pivotal 2012 decision in oncology, the BMS vs. Merck biomarker strategy that ultimately decided the Keytruda-Opdivo war, and found that both Claude and GPT independently recommended the same path BMS took. The path that lost.

The agents produced rigorous analysis, identified the exact competitive threat, and still followed the consensus. As Chang put it: “AI can give you the best possible analysis. It can’t give you the courage to go against it.”

What can AI agents do today, where are they falling short, and why is everyone building them?

This article is for paid subscribers of BiopharmaTrend Pro membership.

Keep Reading -->

Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

Subscribe now

There’s a quick format poll at the bottom — we would appreciate your vote.

🤖 AI x Bio

(AI applications in drug discovery, biotech, and healthcare)

Foundation models and atlases keep spreading:

• Xaira‘s virtual cell model, which we covered last week, got a longer treatment in a Decoding Bio interview where Bo Wang and the team talked through their bet on interventional Perturb-seq data over observational atlases.

• Bioptimus launched a spatial biology ‘STELA’ atlas profiling 100,000 patient tissue specimens across three continents with 10x Genomics and Broad Clinical Labs. Roughly 20x the scale of existing spatial biology datasets then feeds into their biology world model.

• Meta open-sourced TRIBE v2, a brain encoding model trained on 700+ subjects that predicts whole-brain fMRI responses to video, audio, and text at 70x the resolution of its predecessor, functioning as a kind of digital twin for in-silico neuroscience experiments.

• Not strictly a techbio item, but relevant to the atlases and the neurotech thread we covered in our Neurotech Review: a UNC-led team published the first comprehensive atlas of human brain functional connectivity across the full lifespan, drawing on fMRI scans from 3,556 people aged 16 days to 100 years (Nature).

Pushing back on the general mood of enthusiasm, a recent Nature Biotechnology review co-authored by Eric Topol defines “generalist biological AI” (GBAI)—unified systems that interpret, synthesize and scale across several biological domains—and maps both the progress and the persistent gaps.

Accomplishments:

• Language models for nucleotides, proteins, cellular data, and metabolomics (scGPT, Evo 2, ESM-2, DreaMS)

• Structure prediction and design (AlphaFold 3, RoseTTAFold All-Atom, RFdiffusion, Boltz-2, ATOMICA)

• Microscopy and histology image analysis (CellPose, Virchow2, UNI)

• Spatial transcriptomics integration toward virtual cells (Nicheformer, scGPT-spatial, CORAL)

• Early agentic frameworks for autonomous discovery (Virtual Lab, Biomni, SpatialAgent)

Challenges:

• Foundation models are expensive to train, difficult to interpret, prone to hallucinations, and “likely less effective than they appear” given team-selected benchmarks and proof-of-concept evaluations

• Simpler specialized models consistently match or beat foundation models on tasks like gene perturbation prediction, raising the question of whether foundation model development is necessary for marginal improvements

• Context length limits prevent capturing long-range genomic dependencies such as enhancers and epigenetic features

• Joint encoding spaces across modalities remain underdeveloped—incorporating gene expression alongside nucleotides and amino acids is “not nearly as intuitive”

• Biological complexity remains difficult

• Data scarcity for RNA structures, eukaryotic genomes and perturbation response datasets hampers generalization

• Experimental validation is shallow, because most models lack wet-lab validation, and the gap spans in silico/in vitro/in vivo levels, with organoids proposed as a bridge

Path forward: rather than treating foundation models as the default solution, authors argue for selectively deploying specialized models where they outperform, integrating them within agentic workflows that can call domain-specific tools as needed. Combine multimodal datasets (Human Cell Atlas, HuBMAP, HTAN), validate predictions experimentally at scale (including through organoid-based closed-loop systems) and build toward virtual cells that can simulate perturbation responses across biological layers.

📝 An essay in Asimov Press by Alvin Djajadikerta argues that current AI training architectures are structurally locked into what he calls ‘hypernormal science’—not as a side effect of scale, but because systems trained to minimize prediction error against existing datasets cannot derive concepts outside the variables those datasets encode. He points to early empirical evidence: a study of 41 million papers found AI-augmented research covers ~5% less topical ground despite higher output. Basically, AI scientist pipelines face a built-in evaluation problem in that the only available proxy for idea quality is consistency with the current paradigm, which is exactly what paradigm-shifting work violates.

⚡ In brief

🔹 insitro and Bristol Myers Squibb expanded their ALS collaboration with two additional AI-identified targets (ALS-2 and ALS-3), joining the first target BMS nominated in December 2024, all found through insitro’s Virtual Human platform.

🔹 Tempus and Daiichi Sankyo partnered on AI-driven biomarker discovery for an undisclosed ADC program in oncology, using Tempus’ multimodal foundation model for patient stratification and response mapping.

🔹 Iambic Therapeutics disclosed the chemical structure of its brain-penetrant HER2 inhibitor, now in Phase 1b (the program went from inception to IND in two years). Iambic also received a £4.5M compute grant to train the next generation of their protein-ligand structure prediction model.

🔹 Philips got FDA 510(k) clearance for an AI copilot that tracks mitral valve repair devices in real time during minimally invasive heart procedures by fusing echo and fluoroscopy imaging.

🔹 Ataraxis AI launched Breast CTX, a test that uses causal inference to estimate individualized chemotherapy benefit in breast cancer — separating baseline prognosis from treatment effect rather than relying on average group response. Said to be validated on 10k+ patients, already adopted at NCI-designated centers.

🔹 Purple Biotech partnered with Converge Bio to apply generative AI to its CAPTN-3 tri-specific antibody platform for solid tumors.

Brian Buntz at R&D World profiles the Wiley-OpenEvidence deal as a case study in why curated content is becoming a bottleneck for medical AI. Josh Jarrett (Wiley SVP, AI growth) argues the value split between AI technology and trusted source content is closer to 50-50 than most people assume, and that off-the-shelf ‘deep research’ tools search broadly but fail to go deep into proprietary or domain-specific datasets. He also flags a novice-expert gap: generalist AI is most dangerous for users in the middle of the expertise spectrum, who lack the domain knowledge to catch errors but trust the output more than true novices do. Note: Since 2025, Wiley is collaborating with Anthropic on principles for how AI agents should interpret scientific literature: distinguishing preprints from versions of record, handling retractions, and treating published evidence as evolving dialectic rather than settled fact.

A small open-source moment: PhytoVenomics released Blatant-Why, an autonomous antibody design agent for Claude Code that wires together BoltzGen, Tamarind Bio, and Adaptyv Bio into a single target-to-lab-ready-VHH pipeline. Their point is that the agentic orchestration layer is not the hard part—models and APIs are open and available.

💰 Money Flows

(Funding rounds, IPOs, and M&A for startups and smaller companies)

🔹 Biggest deal of the week: Insilico Medicine and Eli Lilly signed a $2.75 billion agreement covering AI-driven discovery of oral small-molecule therapeutics across multiple programs, with Lilly getting exclusive worldwide commercialization rights. Insilico gets $115 million upfront, with the rest in development, regulatory, and commercial milestones plus tiered royalties. Lilly is essentially plugging its own target selection into Insilico’s generative chemistry engine. Lilly picks the biology, Insilico generates the molecules.

The two have been working together since 2023, first through a software licensing arrangement, then a ~$100 million compound-design collaboration in 2025, so this is a third iteration.

Insilico listed on the Hong Kong Stock Exchange on December 30 in a ~$293 million IPO, where Lilly itself was a cornerstone investor. The company now has 28 preclinical candidates nominated since 2021 across over 40 programs, with 12 IND approvals and its lead idiopathic pulmonary fibrosis asset having reported Phase 2a data in Nature Medicine last June. STAT says Insilico’s pipeline page was briefly updated to show a GLP-1 candidate out-licensed to an undisclosed partner, though no annotation is visible there yet. Insilico’s CEO has also mentioned he wants to “develop the next” Mounjaro.

🔹 Merck put $20 million upfront into an AI-enabled target discovery deal with Quotient Therapeutics (Flagship-backed, up to $2.2 billion in milestones) for inflammatory bowel disease. Quotient’s angle is somatic genomics: mapping mutations that accumulate in individual tissues over a lifetime rather than relying on inherited variants. Merck already has a major IBD position after spending nearly $11 billion on Prometheus in 2023.

🔹 Broadly, IQVIA‘s recent 2025 annual numbers painted a split picture: overall biopharma funding down 20% to $82 billion, IPOs at a 10-year low ($3 billion), but mega-deals above $2 billion surging from 27 to 68, worth a combined $360 billion.

🚀 A New Kid on the Block

(Emerging startups with a focus on technology)

🔹 Neion Bio hatched out of stealth this week with a somewhat odd (but not entirely new) premise: genetically engineer chickens so their eggs produce pharmaceutical proteins—the same complex biologics (Humira, Keytruda) that currently require billion-dollar Chinese hamster ovary cell facilities to manufacture. Co-founder Sam Levin says the approach could cut production costs by 10-100x, and that 3,900 hens could cover global Humira demand. They’ve announced a deal to develop three compounds with a major pharma partner they won’t name yet.

The company has 50 engineered roosters so far, and the next generation of hens is where we see whether the birds reliably produce the target proteins at useful concentrations. People have been trying to make pharma-chickens work for 30 years with little to show for it (one FDA-approved egg-produced drug from 2016, Kanuma, is still priced at $310K/year per patient), but Neion says newer primordial germ cell techniques finally make the engineering practical. The more ambitious play is to skip embryo work entirely and use viral gene delivery in adult hens, which Neion’s CSO describes as “gene therapy for chickens.”

🔹 Triangle Health raised $4M in pre-seed for an AI health navigation platform pitched around the agentic AI angle—the idea being that a patient’s full medical record becomes the context layer for AI systems that can research treatments on their behalf. Similar to the Australian dog-cancer story we covered a few issues back, but productized. Origin story: co-founder Arun Verma got a surprise Grade 2 glioma diagnosis from a Prenuvo scan in 2023, had surgery, then spent a year using ChatGPT to research his condition and found two experimental therapies his oncologist agreed to try. He’s now doing a Master’s in AI at JohnsHopkins and built Triangle to replicate that research process for patients who can’t do it themselves.

Share

Read also:

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

Cancer as a Data Problem: What AI Is Doing in Oncology

GTC 2026 brought that energy straight into life sciences. NVIDIA CEO Jensen Huang's keynote framed agentic AI as the next platform shift, and the life sciences sector seems to be taking that literally. Agents are now everywhere you look: IQVIA rolled out over 150 of them across clinical and commercial workflows, Insilico added one for translational biology, Latent Labs has one designing antibodies from text prompts, and NVIDIA itself unveiled a full surgical robotics stack.

We now have two top pharma companies in an apparent compute arms race within the same quarter. Roche's 3,500+ Blackwell GPUs positioned as the largest announced pharma GPU footprint, right on the heels of Lilly‘s LillyPod supercomputer and its $1B co-innovation lab with NVIDIA announced at JP Morgan.

Startups are matching the energy: Xaira, backed by nearly $1 billion in funding, has now launched a 4.9-billion-parameter virtual cell model, Earendil, since our note of their plans for a ~$500M IPO last week, pulled in $787 million for AI-driven biologics, and the overall VC wave into AI-native bio keeps building. Speaking broadly of AI-native startups, Huang pointed to ~$150 billion in venture funding flowing their way last year.

Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

Subscribe now

🤖 AI x Bio

(AI applications in drug discovery, biotech, and healthcare)

🔹 Roche launched its own large-scale AI computing hub for drug discovery, development, and manufacturing after buying 2,176 Blackwell GPUs (bringing its total capacity above 3,500), supporting model training, biological data analysis, lab-linked workflows, and factory digital twins.

🔹 AI reasons about protein function — Arc Institute researchers introduced a multimodal model for protein function prediction that combines protein embeddings with language-model reasoning, beating prior public benchmarks, generating step-by-step biological explanations, and earning preference over curated annotations in 79% of expert comparisons.

🔹 At GTC, UCL and partners unveiled what they describe as a first hybrid quantum-GPU biomolecular simulation pipeline, combining quantum hardware with 120 H100 GPUs to model a major drug-target protein class at realistic biological scale.

🔹 Xaira Therapeutics launched a 4.9-billion-parameter AI virtual cell model that predicts cellular responses to genetic perturbations, trained on 25.6 million perturbed single-cell transcriptomes and reported to generalize to unseen biological contexts.

🔹 Latent Labs, after coming out of stealth last February, launches an AI agent that designs therapeutic antibodies from text prompts in hours, reporting ~67% success with nanomolar affinities.

🔹 At GTC 2026, NVIDIA launched Proteina-Complexa, a generative protein binder design model accepted as an oral presentation at ICLR 2026, built on Proteina (ICLR 2025) and extending it to full atomistic binder design with inference-time compute scaling. Novo Nordisk, Manifold Bio, and Viva Biotech participated in wet lab validation. Other announcements also included a surgical robotics stack: Open-H (700+ hours of surgical video), Cosmos-H (synthetic surgical data generation), GR00T-H (vision-language-action model for clinical robotics), and Rheo (a simulation framework).

🔹 IQVIA launched a unified AI platform to support clinical, commercial, and real-world decision-making, with 150+ deployed agents, and 100+ AI patent filings.

🔹 Open-source protein design workflow — Dyno Therapeutics unveiled an open-source agentic AI suite for protein binder design that pairs generative models with lab-trained predictive filters.

🔹 Agentic AI for translational biology — Insilico Medicine launched a new AI analysis agent for therapeutic discovery that lets biologists use natural language to run multi-omics and bioinformatics workflows, drawing on 140+ scientific skills and 1,000+ tools.

🔹 To fix the experimental bottleneck, Arctoris opened a new biophysics center in the UK to generate drug-binding and protein-quality data at scale for AI-driven discovery.

🔹 Building a massive genomics dataset for AI — Basecamp Research launched a Trillion Gene Atlas to collect genomic data from 100+ million species, aiming to expand biological training data for AI models through global sampling.

🔹 AlphaFold adds protein interaction context — the AF database now includes 1.7 million predicted homodimer structures from an initial 30 million candidates, expanding beyond single proteins to capture how identical protein pairs may interact across 20 key organisms.

🔹 Automated labels can quietly worsen medical AI bias — Researchers found breast tumor segmentation models do far worse on younger patients for qualitative reasons, and training on machine-generated labels can amplify the disparity by ~40% while biased benchmarks can hide the damage.

💰 Money Flows

(Funding rounds, IPOs, and M&A for startups and smaller companies)

🔹 Earendil Labs raised $787 million to scale AI-driven antibody and biologics discovery, expand teams, and advance a pipeline of 40+ programs, including one program ready for Phase 2 and multiple regulatory filings planned for 2026-2027.

🔹 Neuracle‘s commercial approval triggered a funding wave across China’s BCI sector: Shanghai-based StairMed raised ~$73M for its robot-inserted flexible electrode approach, while Gestala pulled in $21M just two months after launch for a non-invasive ultrasound BCI platform.

🔹 Verily raised $300M and became an independent company, using the capital to expand its AI-ready health data platform, consumer health tools, and clinical/research agents while Alphabet shifts to a minority stake.

🔹 Rivia raised €13M to build AI agents for clinical trial operations, expanding its unified data platform to proactively flag risks, anomalies, and enrollment issues in regulated trial workflows.

⚙️ Other Tech

(Innovations across quantum computing, BCIs, gene editing, and more)

🔹 Nia Therapeutics received FDA Breakthrough Device status for an AI-guided brain implant aimed at memory loss after traumatic brain injury, with a study reporting a 19% recall improvement by detecting poor memory-encoding states and delivering targeted stimulation.

🔹 Brain activity restored after deep freezing — researchers in Germany revived key functions in vitrified mouse brain slices after storage at up to −150 °C for as long as seven days, preserving neuronal firing, metabolism, and synaptic plasticity.

🔹 ReVision Implant won FDA Breakthrough Device status for a brain-stimulation visual prosthesis designed to restore functional vision in blindness beyond the retina, with planned surgical testing in October 2026.

🔹 Mind typing nears everyday speeds — Massachusetts General Hospital researchers enabled two people with paralysis to type on a standard digital keyboard using brain implants that decode imagined finger movements, reaching 22 words per minute with low error after training on just 30 sentences.

🔹 “No, we haven’t uploaded a fly yet”—scientists push back on viral claims of a digital fruit fly, arguing the demo stitched together existing brain and body models with pre-programmed behaviors, falling far short of true whole-brain emulation.

🏛️ Bioeconomy & Society

(News on centers, regulatory updates, and broader biotech ecosystem developments)

🔹 FDA pushes shift beyond animal testing, issuing draft guidance to validate non-animal methods like organoids, lab models, and simulations for drug safety testing, aiming to replace animal studies with more predictive human-relevant data and speed clinical development.

🔹 NIH committed $150M investment in human-based research to reduce reliance on animal models, framed around its Complement-ARIE program.

🔹 arXiv’s move out of Cornell—Science reports it will become an independent nonprofit on July 1, 2026.

🚀 A New Kid on the Block

(Emerging startups with a focus on technology)

Former GSK AI Leaders Launch Stealth Oncology Startup — Shane Lewin and Nick Peterson, who both recently left GSK after six years building the pharma giant’s AI/ML and research data infrastructure, have co-founded Clockwork Bio (Lewin as CEO, Peterson as CTO).

The stealth-mode startup is building an AI-native platform for preclinical oncology drug discovery, arguing that conventional pipelines designed around heritable genetic targets fail for cancer, where disease is driven by epigenetic regulation, cell signaling, and lineage switching.

Their approach combines active learning, AI-based phenotyping, and emerging modalities to force cells between healthy and diseased states in vitro while learning the underlying biology at scale.

No funding or team details disclosed, formal product launch expected later in 2026.

Share

Read also:

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

Cancer as a Data Problem: What AI Is Doing in Oncology

Caveat: the science of personalized oncogenomics isn't new, and the AI mostly helped a non-specialist navigate tools that already exist. This still required serious money and a lab willing to help, but the tooling is getting faster and more navigable even for non-specialists.

Meanwhile, Google’s diagnostic AI ran pre-appointment patient interviews at a real clinic—100 patients, no safety incidents, accuracy allegedly on par with primary care doctors. OpenFold Consortium now released its full training pipeline with datasets, weights, and training code.

On the money side: an AI-driven protein engineering startup Earendil is reportedly eyeing a Hong Kong IPO that could raise up to $500M. Breakout Ventures closed $114M for early-stage neuroscience and biomedical AI. Seemay Chou launched Radial, a $500M effort aimed at the scientific data infrastructure that AI tools depend on but nobody wants to fund. And Owkin spun out its diagnostics work as Waiv with $33M for oncology stratification.

Elsewhere: China cleared the first minimally invasive BCI for everyday home use. Researchers published a molecule-level simulation of a full bacterial cell cycle. Also—arXiv is separating from Cornell to go independent and is looking for a CEO.

Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.

If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:

Subscribe now

🤖 AI x Bio

(AI applications in drug discovery, biotech, and healthcare)

🔹 A pet owner used AI to build his dog a custom cancer vaccine and it worked. After surgery and chemo failed, an Australian engineer used ChatGPT to map a treatment plan, AlphaFold to identify tumor targets, and eventually convinced a university lab to manufacture a personalized mRNA vaccine in under two months.

🔹 Microsoft launched Copilot Health, a dedicated health companion that pulls together wearable data, medical records from 50,000+ US providers, and lab results into one place.

🔹 An AI doctor’s assistant just passed its first real-world test. Google‘s diagnostic AI conducted pre-appointment patient interviews at a real clinic—100 patients, zero safety incidents, and its diagnostic accuracy matched that of primary care doctors.

🔹 The OpenFold Consortium released the complete OpenFold3 training pipeline—datasets, weights, training and inference code—making this biomolecular complex prediction system fully reproducible and extensible and not just inference-accessible. Antibody-antigen prediction remains an open frontier and a stated 2026 priority.

🔹 AI beats specialists at rare disease diagnosis. A Nature-published agentic system integrating 40+ clinical and genomic tools outperformed experienced physicians, with traceable, evidence-linked reasoning; over 500 institutions signed up since its July 2025 launch.

🔹 Structural confidence scores don’t predict cellular function. A 12,000-design CAR-T binder benchmark found only 5.9% of AI-generated designs worked (rising to 10.6% with basic sequence filters) while standard metrics (pLDDT, ipTM, Rosetta) had near-zero predictive power. The core failure: RFdiffusion’s helix-heavy backbones combined with ProteinMPNN’s bias toward lysine/glutamate produce sequences that stall ribosomes in cells. The one top-performing team folded designs in full CAR context rather than isolation.

🔹 Weeks of compute compressed into hours. Schrödinger says it uses Google Cloud GPUs to screen billions of compounds in a single weekend.

🔹 A living cell, simulated molecule by molecule. Researchers built a complete 4D whole-cell model of a minimal bacterium covering gene expression, DNA replication, metabolism, and division across a ~100-minute cell cycle. Each simulated cell is unique due to stochasticity, and the model correctly predicts doubling time, ribosome counts, and daughter-cell variability.

🔹 Asimov Press published a beginner-accessible walkthrough on computational antibody design using BoltzGen.

🔹 BigHat Biosciences benchmarked 30+ protein language models for antibody design with wet lab validation and found that model choice matters less than how one uses it, and that even a few hundred experimental datapoints outperforms any zero-shot approach.

💰 Money Flows

(Funding rounds, IPOs, and M&A for startups and smaller companies)

🔹 AI drug discovery startup Earendil (combining generative protein engineering with wet lab work to develop biologics) is reportedly eyeing a Hong Kong IPO that could raise up to $500M, following Insilico Medicine‘s successful listing there last year. Partnerships with Sanofi, Wuxi XDC, and Wuxi Biologics already in place.

🔹 Breakout Ventures raised $114M for its third fund, targeting early-stage neuroscience and biomedical AI: lab automation, computational chemistry, and AI-powered research tools. Early bets include a lab robotics startup; their track record includes a $300M Halozyme acquisition and the first FDA-cleared rapid sepsis diagnostic.

🔹 DeepMind and MIT alumni raised $13.5M to build a platform that connects ideas across scientific fields, proposes testable hypotheses, and runs simulations before experiments.

🔹 Using salmonella to fight cancer. Salspera is raising $91M to fund a phase 3 trial of an engineered salmonella strain that delivers an immune-boosting signal directly into tumors, previously showing reduced tumor burden and improved survival in metastatic pancreatic cancer. A two-person team, an orphan drug tag, and a Nasdaq listing on the horizon.

⚙️ Other Tech

(Innovations across quantum computing, BCIs, gene editing, and more)

🔹 First invasive BCI for paralysis cleared for everyday use in China—a minimally invasive implant that sits above the motor cortex has been cleared for everyday home use in paralyzed patients, translating imagined hand movements into a robotic glove.

🔹 Large gene insertion without the toxicity. Massachusetts General Hospital researchers developed a genome editing approach using circular single-stranded DNA donors to achieve kilobase-scale gene integration, avoiding the immune response triggered by the double-stranded DNA donors traditionally required for large insertions.

🏛️ Bioeconomy & Society

(News on centers, regulatory updates, and broader biotech ecosystem developments)

🔹 The internet’s preprint library is going independent. arXiv—the free platform hosting 2.7 million scientific papers that the entire AI and life sciences research world depends on—is separating from Cornell University to become an independent nonprofit, and is hiring its first CEO at ~$300K. The academic community’s main worry is that independence means paywalls eventually follow.

🔹 UK bets £500M on homegrown AI. The government’s new Sovereign AI fund launches April 16, with life sciences as a priority sector.

🔹 A Middle Eastern shipping lane could empty American pharmacy shelves. Nearly half of U.S. generic prescriptions come from India, which depends on the Strait of Hormuz for 40% of its crude oil—a key input for drug manufacturing. Current stockpiles give 30–60 days of buffer, but a prolonged closure would hit generics hardest, as thin margins leave little room to absorb rising fuel and petrochemical costs.

🚀 A New Kid on the Block

(Emerging startups with a focus on technology)

🔹 Owkin spun out its diagnostics arm as Waiv with $33M in fresh funding, focused on AI-powered patient stratification tests for oncology—both in clinics and clinical trials.

🔹 AI in science is only as good as the infrastructure beneath it. Researcher Seemay Chou launched Radial, a $500M organization focused on modernizing how scientific data is generated, shared, and built upon—the “unglamorous” infreastucture and tools AI needs to deliver real value. Early projects include better methods for capturing protein dynamics.

Subscribe now

Read also:

Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape

AI Protein Design: Platforms, Specialists, Modular Tools

The ‘Holy Grail’ of Digital Biology

⚠️ The most immediate one is patient recruitment. Far back in 1979, the father of clinical pharmacology Louis Lasagna observed that the pool of eligible patients shrinks by 90% the moment a trial opens, only to reappear once it closes. Lasagna’s Law remains as relevant as ever: according to a 2022 article, 11% of trial sites enrol zero participants and nearly 90% of trials face meaningful delays. With Phase II and III trials costing roughly $40,000 per day, the financial toll is brutal.

⚠️ Another issue is clinical attrition. Research from VU Amsterdam found that between 2012 and 2019, the share of trials successfully completing each phase declined steadily—particularly at Phase II. In the first half of 2024, nearly a third (32%) of trials were discontinued at Phase II—a 56% rise compared to pre-pandemic levels. Combined with stagnant rates of Phase III initiation over that same decade, the picture is one of a systemic bottleneck: trials that begin are increasingly unlikely to see the finish line.

⚠️ Rare disease research presents its own distinct challenge. As the FDA’s Rare Disease Evidence Principles note, shrinking patient populations make it progressively harder to generate reliable efficacy data through conventional designs—especially placebo-controlled trials, where enrolling enough participants to reach statistical significance can be close to impossible.

⚠️ Apart from operational intricacies, there is the ethical dilemma. Randomized controlled trials remain the gold standard for evaluating new therapies, but randomization isn’t always defensible. When an effective treatment already exists, assigning patients to a placebo raises serious moral questions, e.g. HIV cure trials with the antiretroviral treatment interruption.

The question, then, is whether parts of the control process can be simulated rather than physically recruited.

Digital twins are emerging as a compelling response. Last October, Sanofi Ventures invested in a digital twin platform developer QuantHealth, bringing its total funding to $30M. In 2025, the FDA announced plans to phase out animal testing requirements for monoclonal antibodies in favor of human-relevant methods, including AI-driven computational models—with the EMA moving in the same direction. Both industry and regulators, it seems, are taking this technology seriously.

👥 How it Works

A digital twin is a virtual replica of a physical object, continuously updated with real-world data so it mirrors the original’s behavior in real time. The concept, first applied by NASA in the 1960s, has since migrated from engineering into healthcare.

The applications are wide-ranging: optimizing industrial processes as Eli Lilly did to boost production of their GLP-1 drugs, predicting equipment failures, streamlining supply chains, and accelerating product development.

In a clinical trial patients are generally divided into two groups, also known as arms. The intervention arm receives the experimental treatment; the control arm receives a placebo, standard-of-care treatment or sham, serving as the baseline against which results are measured. Randomized controlled trials (RCTs) are the gold standard because randomization minimizes bias, but that randomization isn’t always flawless.

The traditional workaround of external controls drawn from historical trials, health records, or registries all carry their own limitations. For instance, data like those don’t include underrepresented groups or don’t account for placebo effect due to their observational nature.

Digital twins go a step further: using AI models augmented with historical data, they generate individualized predictions of how a patient might respond under different treatment scenarios. When used to simulate outcomes for patients who do not receive the experimental therapy, these models can produce a synthetic control arm.

These trial-level twins build on a foundation of patient-specific digital twin modeling (virtual replicas of individual physiology shaped by genomics, imaging, and clinical history) which we covered in our earlier overview of biological and patient-specific twins.

In clinical trials, an AI-powered digital twin typically operates in three steps:

This article is for paid subscribers of BiopharmaTrend Pro membership.

Keep Reading -->