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:

→ Proprietary data generated in-house (e.g., Recursion's phenomics screens, Insilico's chemistry-plus-clinical loop, Owkin's federated hospital access, etc).

Anthropic doesn't seem to have this yet. It rents it (e.g., Basecamp's BaseData, PacBio/Ultima sequencing in the Trillion Gene Atlas). The players who own their data-generation flywheel keep that edge (for now).

→ Wet-lab-to-model feedback loops. The ones who've vertically integrated experiment and prediction still have the tight loop Anthropic says it's trying to build by running its own program.

→ Clinical pipelines, optionality to pivot programs fast, for those who do have a pipeline.

Where Anthropic does compete with them, and this might be the real pressure, is on general reasoning as commodity infrastructure.

If Claude Science does a considerable % of what a top/mid-tier computational biology platform can do, available to every paid subscriber, the AI-biotechs lose their "we have special AI" pitch to pharma. Their story has to become "we have special data and validation," which is a harder, more capital-intensive sell, but leading players are doing it alright.

That being said, I do see real pressure from Anthropic’s competition for the following situations:

1) new AI drug discovery companies

2) pureplay companies that sell only AI-enabled research tools or services.

Those will have to build a defensible moat real quick.

Leave a comment

The Numbers Behind China’s Biopharma Ascent

A new comprehensive report by Information Technology and Innovation Foundation (ITIF), a top US think tank for policymaking, makes the case that China is on track to challenge U.S. leadership in biopharma within a decade. The report is a policy argument, urging a U.S. response, but here let’s focus on the underlying data.

First of all, what is interesting is approval time for human trials in China fell from 501 days to 87. Patient enrollment at Chinese tertiary hospitals runs 5–10x higher than at U.S. academic medical centers, and per-patient costs are lower—a Phase I trial there is about 43% cheaper and over 50% shorter than in the U.S.

Net effect: Chinese firms can go from discovery to first-in-human in roughly half the global average time. Speed and cost, not just science, are the edge.

The output numbers have caught up to the process. China’s share of clinical trials for the most innovative drugs jumped sixfold in a decade to 30%, nearly level with the U.S. at 33%.

It now accounts for 31% of the global drug pipeline. Value-added pharmaceutical output grew roughly 13-fold since 2002, to ~$187 billion.

Three Chinese firms now sit among the world’s 20 largest pharma pipelines, led by Jiangsu Hengrui Pharmaceuticals at #12.

China has also overtaken the U.S. in highly cited biotech publications—nearly 800 top-decile papers in 2024 versus 112 for the U.S.

The deal wave is where Western pharma is voting with real money, though.

Out-licensing deals from China grew 31x since 2015 (5 to 157 deals), with value up 54x to $135.7 billion last year—and another $60 billion in Q1 2026 alone. Average upfront payments more than tripled to $172 million.

The marquee examples: AstraZeneca paid CSPC $1.2 billion upfront for weight-loss assets; Bristol Myers Squibb committed up to $15.2 billion to Hengrui across 13 programs; Pfizer struck multibillion-dollar oncology deals with 3SBio and Innovent... and so on.

China is now the leader in specific frontier modalities too—over 42% of the global antibody-drug-conjugate pipeline, and it overtook the U.S. in oncology research output in 2024.

The report is measured about the limits, though.

China still holds only ~4.8% of the global biotech market and ~7.5% of global pharma sales. Much of the ecosystem rests on state subsidies (one study found 99% of China’s top R&D-spending firms received them), and FDA reluctance to approve drugs on China-only trial data remains a barrier to reaching Western markets.

So, the lead is in inputs and velocity more than in commercialized, globally approved products—for now...

Also glad to see BioPharmaTrend cited in the report.

Anyway, worth a read for anyone tracking where drug development is moving in China and how it affects the global life science market.

I spent last week in Amsterdam, reporting for BiopharmaTrend at 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.

The Rise of “Verification layer”

There is an increasing number of companies that are positioning healthcare AI tools/services not as a producer of clinical content/material, but as a means to verify it.

For instance, Berlin-based aiomics has put into production a verification layer that sits on top of hospital IT systems, turning the unstructured documents a hospital receives (e.g. faxes, referrals, scans, dictation) into a structured, sourced patient record. Rather than generating clinical text and hoping it is correct, the system audits every statement against the original source document through a multi-agent protocol — the company’s answer to what it calls the central risk of generative AI in medicine: fluent, plausible output built on bad data. It is live across more than 30 hospital sites in Germany, certified to ISO 27001, and is being independently evaluated at the Charité in Berlin.

The same positioning showed up elsewhere. Guideways AI is launching EU MDR Reviewer and QMS Reviewer, putting AI on the reviewer’s side of a full CE-certification submission to catch issues before they cause delays. AMBOSS pointed to a benchmark result: its clinical AI search agent, LiSA, ranked first overall and in the top safety tier of the NOHARM study from Stanford, Harvard and collaborators, which assessed 31 AI systems across 100 real clinical scenarios using 12,747 expert annotations — a result the company credits to drawing only from curated clinical sources. And FiveBrane’s Datametior turns the lens on the data itself, scoring how AI-ready a dataset is before anyone spends money training on it.

Share

“Human-in-the-loop” is probably a product category, not a caveat

The phrase recurred often enough that it read less like a reassurance and more like a product. Kimberly Noel, Roche’s Global Lead of AI Advocacy and Digital Health, pointed to Human-In-the-Loop GmbH, a company built entirely around the idea. Its stated sole mission is promoting AI transformation while staying compliant with the EU AI Act and ISO/IEC 42001, and it argues that “human in the loop” is not a metaphor but the operating model.

The same language showed up at the primary-care practice Hausärzte am Spritzenhaus, which runs AI-based task steering under what it calls a strict human-in-the-loop architecture, and at Longevity AI, whose Florence 2.0 is built to keep the doctor at the center of every patient interaction.

EU digital sovereignty as a selling point

Among the European companies, regulatory readiness and data sovereignty kept appearing as selling points rather than mere “disclaimers”. Datum Agent positioned itself as the first vertically integrated, EU-sovereign AI platform purpose-built for healthcare — GPU infrastructure hosted in the EU, patient data processed within EU jurisdiction, and, in its words, “without dependency on US hyperscalers.”

Already mentioned earlier aiomics made a point of running entirely within the EU and holding ISO 27001 certification. iCure said its Cardinal v2 is prepared for NIS-2, the EU AI Act and the European Health Data Space… and so on.

Triage and navigation

Some of the more detailed numbers came from tools that route patients to the right level of care.

Infermedica, with Healthdirect Australia, published peer-reviewed research in Mayo Clinic Proceedings: Digital Health analyzing more than 1.55 million real-world virtual triage interactions, reporting emergency-department intent down from 36.7% to 24.6%, engagement with lower-acuity care more than doubled, and patient uncertainty about where to seek care down 99.6%.

A study published in Nature Health of Ada Health’s integration into South Africa’s MomConnect platform reported that, among 968 participants, the share seeking care more than doubled from 17% to 43%, with recommendations rated safe by an independent physician panel in 98% of cases.

And Tucuvi reported that its clinical voice agent, LOLA, was associated with a 43.7% reduction in urgent COPD admissions and a 40% drop in hospital stays at Hospital Ribera Povisa.

The longevity is becoming a mainstream term

Longevity was hard to miss. For instance, Longevity AI announced launching an AI platform for preventive-care practices, built on more than 1.6 million longitudinal health records and used by systems including Maccabi and Clalit.

Reya.ai, a 2026 Health 2.0 Award winner and NVIDIA Inception member, announced Reya Essentials to lower the barrier to entry for new longevity clinics.

Beyond the crowded field of GLP-1 companion apps, Lumen made the case for metabolic intelligence as the next step after GLP-1’s effect on obesity treatment — tracking how the body responds to the drugs over time to support engagement and long-term outcomes, drawing on more than 100 million measurements from over 350,000 users.

Unfiltered released its Longevity 100 power list as part of an investment report, and Kearney and Microsoft put out a report on technology and longevity alongside a panel asking, pointedly, whether longevity is just prevention rebranded.

Adherence and persistence

Medication non-adherence kept surfacing as a problem people had put numbers to.

BrightInsight, with Sanofi, tracked more than 6,000 specialty patients and presented real-world evidence on persistence at 12 months, opening on the figure that 71% of specialty patients abandon therapy within a year.

PACE Clinical, which is integrating the BEAMER project’s B-COMPASS model, cited EU figures of 200,000 premature deaths a year and over €125 billion in avoidable healthcare costs tied to non-adherence.

Observia’s SPUR behavioral diagnostic, which the company says is validated by seven publications, is being built into an explainable-AI adherence agent. And Redcare Pharmacy’s smartpatient was mentioned to be working with UCB to support patients with Hidradenitis Suppurativa across the care journey.

Data as a healthcare and pharma asset was the biggest topic

The throughline beneath the AI was data itself. Lots of companies presented solutions in the data infrastructure and data aggregation areas. For instance, Lumen pointed to more than 100 million real-world metabolic measurements from over 350,000 users; PAICON’s PaiX Navigator offered access to disease datasets spanning 60-plus countries, pitched explicitly at closing the representation gap in medical AI; and Leumit opened 23 years of clinical EMR community-care data to European innovators.

Others focused on making existing data usable. Like, Data4Life’s Data2Evidence platform went live at Mount Sinai, opening secure access to more than 12.4 million de-identified patient records, while Briya launched a no-code environment for running real-world evidence and epidemiological studies on clinical data.

A panel on data as pharma’s biggest asset, featuring speakers from AstraZeneca, Roche, Memorial Sloan Kettering, and Germany’s Health Data Lab, broadly agreed on why so much of it stays locked up: the obstacles are less technical than about governance and trust, and bolting AI onto broken interoperability mostly just surfaces the breakage faster.

One speaker noted that the patient is still the one stitching their records together, re-explaining their history to a new healthcare provider roughly 40% of the time. Others pointed to what connected data makes possible — using a combination of large language models to read entire patient charts ahead of a visit and flag likely trial eligibility. One panelist cited a national claims dataset of some 75 million people, opened to research after a change in German law.

Subscribe now

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. Co-folding generates structural hypotheses from sequence alone. Physics-based methods provide rigorous validation. Neither works well in isolation.

The companies doing interesting work at this interface — SandboxAQ, Genesis Molecular AI, Iambic, Schordinger, etc., seem to get this. The real progress is hybrid: learning-based generation feeding into physics-based refinement.

I think the current hype around general-purpose AI agents obscures something important about the pharma and biotech realm of AI progress. In drug discovery, binding is fundamentally a physics problem. Models trained on data can approximate physics, but they can’t replace it, not yet. The teams investing in both sides are the ones to watch.

Now, since conferences were specifically focused on FEP and co-folding methods, I decided to share a couple of trends in those areas here:

Observations about free energy perturbation methods

FEP has become the workhorse for binding prediction. Free energy perturbation calculations can now reliably predict how well a molecule binds to a protein target. What used to require supercomputers and deep specialist knowledge now runs on a few GPUs, thanks to better hardware and better classical force field parameterization.

The shift is from artisanal to industrial. Drug discovery moved from hand-designed molecules to combinatorial libraries in the late ‘90s, and FEP is following the same trajectory, from carefully hand-tweaked individual calculations to routine bulk triaging of large compound sets before synthesis.

The binding problem is increasingly solved; everything else is not. FEP handles potency prediction relatively well, but druglikeness, metabolism, PK, toxicity, and crystal polymorphs remain poorly amenable to computation. So FEP doesn’t replace medicinal chemistry judgment yet; it removes one major bottleneck while the others persist.

The generative chemistry + FEP synergy is the frontier, but it’s hard. AI-generated molecules tend to be structurally diverse (not congeneric series), which means you need absolute binding free energy (ABFE) calculations rather than relative ones. ABFE is more expensive and less accurate. That’s the current bottleneck for combining generative AI with physics-based validation at scale.

Ease of use matters for adoption. The article argues (via Cresset’s Flare product) that automation, error-checking, and cloud access are what turn a specialist method into an everyday tool across organizations.

Here is what Dmitry Lupyan, Research Leader at Schrodinger, got to say about the current state of FEP:

Observations about co-folding methods

Orthosteric binding works reasonably well; allosteric does not, yet. Co-folding methods reliably place ligands in the main (orthosteric) binding site but consistently fail to find allosteric pockets, instead defaulting to the orthosteric site. This is a training data bias problem because orthosteric sites arguably dominate the protein data bank (PDB).

Boltz-1x is the chemical validity winner. Only 1.5% of its predicted ligands had any PoseBusters issue (default settings), versus 56% for Boltz-1, 93% for NeuralPLexer, and 85% for RoseTTAFold. Under stricter criteria, everything degrades significantly.

Prevalence in training data correlates with success. When allosteric sites are well-represented in the PDB (like GCK), predictions improve. But it’s not the whole story — some well-represented allosteric sites still fail.

The dual-ligand trick helps but does not solve the problem. Submitting two copies of the allosteric ligand improved sampling (50% placed correctly), but you still can’t reliably distinguish the correct pose from incorrect ones without external scoring.

The core tension with co-folding methods: These methods show promise as potential replacements for docking and even FEP, but they currently lack physics-based scoring, produce ensembles of unknown quality, and need significant post-processing before they’re useful for prospective drug design.

📢 Announcement: I’ll be at HLTH Europe in Amsterdam this June (15-18) as an invited journalist/science writer. It is arguably Europe’s largest healthcare tech event, with 5,000+ attendees, one in three at the C-suite level.

I’ll be covering what’s actually being said in the hallways, not just on the stages. If you’re attending, let me know, happy to connect in person! And if there’s a specific topic or company you would want me to dig into while I’m there, drop it below.

Subscribe now

Company Picks

During the event, I talked to several company reps and founders, including , and so I decided to summarize some of the interesting companies in this space:

SandboxAQ

It is an enterprise AI company spun out of Alphabet in 2022, currently valued at $5.75B after raising ~$950M. Their drug discovery division, AQBioSim, combines generative AI with physics-based molecular simulation — what the company calls Large Quantitative Models (LQMs).

The key technical claim is their proprietary Absolute Free Energy Perturbation method (AQ-FEP), which, according to SandboxAQ predicts binding affinities without requiring reference compounds, making it applicable to structurally diverse libraries rather than just congeneric series.

The company says it can profile over 20,000 ligands per day at scale using this approach. They report partnerships with AstraZeneca, Sanofi, and UCSF, among others.

Genesis Molecular AI

Founded in 2019 in California (originally as Genesis Therapeutics, rebranded to reflect the AI focus). The company's core platform is GEMS (Genesis Exploration of Molecular Space), which, according to Genesis, combines proprietary deep learning models with physics-based molecular simulation for small molecule drug design.

Their flagship model, Pearl, is a 3D diffusion foundation model for protein-ligand structure prediction that the company claims outperforms AlphaFold 3 on binding pose prediction, notably, trained using large-scale synthetic data generated from physics simulations, not just experimental PDB structures.

Genesis has raised over $300M from investors including a16z, NVIDIA, Fidelity, and BlackRock. They report active collaborations with Gilead, Eli Lilly, and Incyte — the Incyte partnership was recently expanded to cover at least five additional targets, with Incyte sharing proprietary experimental data to further train GEMS. Nate Gruver from Genesis presented at the CoFold Summit in Session 2 on predicting properties beyond structure.

Iambic Therapeutics

Founded in 2019, headquartered in San Diego. A clinical-stage company whose platform combines two main proprietary AI models: NeuralPLexer, a co-folding model for predicting protein-ligand complex structures directly from sequence, and Enchant, a multimodal transformer that according to the company predicts clinical and preclinical endpoints from small, noisy datasets. The company describes its approach as physics-informed — integrating physical principles into AI architectures to improve data efficiency and enable broader exploration of chemical space. Iambic claims to complete full design-make-test cycles on a weekly cadence through tight integration of AI-generated designs with automated high-throughput chemistry and biology. They report their lead oncology program went from program start to clinic in under 24 months. Partnerships include a multi-year collaboration with Takeda announced in early 2026 (potentially worth over $1.7B in milestone payments) and a technology collaboration with Revolution Medicines. Matt Wellborn from Iambic presented at the CoFold Summit

Nostrum Biodiscovery

Founded in 2015 in Barcelona as a joint spin-off of the Barcelona Supercomputing Center (BSC) and the Institute for Research in Biomedicine (IRB Barcelona), with participation from the University of Barcelona and ICREA. Co-founded by Victor Guallar, Modesto Orozco, and Robert Soliva.

The company's core technology is PELE (Protein Energy Landscape Exploration), a Monte Carlo-based molecular modeling algorithm for protein-ligand docking, binding site prediction, and protein surface exploration. Their commercial platform, NostrumSuite, integrates PELE with AI-driven molecular modeling for virtual screening, hit-to-lead optimization, and applications across small molecules, antibody design, targeted protein degradation, and nucleic acid therapeutics.

According to the company, their ALScreen platform combines AI and molecular modeling for virtual screening of both predefined and ultra-large compound libraries. Nostrum describes itself as bridging physics-based simulation and AI — notably, they are rooted in HPC and biophysical simulation rather than coming from the deep learning side.

Apheris

A Berlin-based company co-founded by Robin Röhm that provides federated computing infrastructure for drug discovery. The core premise is that pharma companies hold proprietary structural and molecular data they can't share due to IP constraints, but that data is exactly what co-folding and ADMET models need to improve.

Apheris claims to solve this by bringing computation to the data rather than moving data, enabling multiple organizations to collaboratively train and benchmark AI models without exposing proprietary datasets.

They provide the technology layer for the AI Structural Biology (AISB) Network, an industry-led collaboration that, according to the company, includes AbbVie, Astex, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Genentech, Johnson & Johnson, Sanofi, and Takeda. One of the network's flagship projects is fine-tuning OpenFold3 on proprietary structural data from multiple pharma companies — without that data leaving each organization — in collaboration with Mohammed AlQuraishi's lab at Columbia.

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

💊 The Logic of Drug Rescue

Before we profile the companies, it’s worth understanding why this thesis is surfacing now and why it’s distinct from what came before.

Drug repurposing has a long history. Sildenafil started as a cardiovascular drug before becoming Viagra. Thalidomide was rehabilitated remarkably, decades after its teratogenic disaster in the 1960s, as a treatment for multiple myeloma. These are cases where an approved (or previously studied) molecule found a genuinely new indication.

What companies like a newcover Biossil, as well as more established players like Lantern Pharma, Pathos AI, and BPGbio, are doing is different. They are not necessarily changing the target disease, but the target patient. The hypothesis: within a trial population that produced a negative aggregate result, there are subgroups of patients who responded, and whose response was masked by the statistical noise of everyone who didn’t.

This isn’t a new idea conceptually. Post-hoc subgroup analysis has been part of clinical trials for decades. What’s new is the scale and sophistication of the AI being applied: multimodal foundation models trained on hundreds of petabytes of data, causal inference engines, spatial transcriptomics paired with pathology imaging, and multi-agent systems reasoning across publications and biomarker data.

The question is whether the analytical tools have finally caught up to the biological complexity… or whether we’re just building fancier ways to p-hack.

Subscribe now

🗺️ The Landscape: Who’s Doing What

The companies working in this space share a thesis but diverge significantly in their technical approaches, therapeutic focus, and maturity. Here’s how the landscape breaks down.

⭐ Biossil

The freshest entrant. Biossil emerged from stealth with $70M co-led by Founders Fund and OpenAI, and a portfolio of ten molecules acquired quietly over three years. Their approach centers on reanalyzing late-stage clinical failures to identify patient subgroups where a meaningful treatment signal was hidden by population heterogeneity. Trials are running across glioblastoma, Alzheimer’s, and other indications.

What makes Biossil notable is the breadth of their bet: ten molecules across multiple therapeutic areas, funded by investors who have not traditionally played in biopharma. The OpenAI connection signals a belief that general-purpose AI capabilities, not just domain-specific biostatistics, can crack the patient stratification problem. That’s an interesting bet, though one that remains unproven clinically.

Details on their technical platform are still limited. We’ll be watching for specifics on what data they’re training on, how their models identify subgroups, and, most critically, whether their approach produces prospectively validated biomarkers or just retrospective correlations.

⭐ Lantern Pharma (Nasdaq: LTRN)

Lantern has been working on a related playbook for years, making it one of the most useful reference points for whether the thesis actually holds up in the clinic. Their RADR AI platform identifies abandoned clinical-stage drugs and matches them to patient subgroups most likely to respond. The focus is oncology.

The most tangible proof point right now is LP-300, a candidate in development for never-smoker non-small cell lung cancer (NSCLC) — a molecularly distinct type of the disease with poor outcomes and no approved therapies focused on this specific population. Lantern has just announced that the FDA raised no objections to key proposed protocol amendments for the Phase 2 HARMONIC trial, a meaningful de-risking step.

Three protocol changes are worth noting:

• Focused enrollment on the EGFR exon 21 L858R subgroup — the molecular subset, accounting for roughly 40% of EGFR-mutant NSCLC globally, where current therapies leave the largest unmet need, according to the company, and where LP-300’s preliminary data have been most differentiated.

• Extended dosing from a maximum of 6 to 8 cycles.

• Transition from a randomized to a single-arm design — intended to accelerate enrollment and sharpen the clinical signal in a genomically defined subgroup.

On the AI side, Lantern has been developing withZeta.ai, a multi-agentic system derived from RADR that extends the platform’s mechanistic modeling capabilities. For LP-300, withZeta has been used to interrogate the drug’s mechanistic potential in the L858R setting — reasoning across publications and biomarker observations to surface insights that informed the development strategy. It’s an interesting example of AI agents being used not just for patient selection but for mechanistic hypothesis generation.

⭐ Pathos AI

Pathos is arguably the company with the deepest data moat in this space, thanks to its roots in the Tempus ecosystem. Founded by executives from Tempus (Eric Lefkofsky’s healthcare AI company), Pathos claims to have access to over 200 petabytes of multimodal oncology data linked to patient outcomes, reportedly 50 times the size of The Cancer Genome Atlas, the largest public genomic dataset in oncology.

Their thesis mirrors Biossil’s: drugs fail because they were tested in the wrong patients, with the wrong assumptions, in trials that couldn’t answer the real question “who benefits, and why?” But Pathos is focused exclusively on oncology and is building what it describes as the largest foundation model in the field.

In April 2025, Pathos entered a major three-way collaboration with AstraZeneca and Tempus to build a multimodal oncology foundation model, with $200 million in data licensing and model development fees flowing to Tempus. The foundation model is being built on Tempus's repository, which includes 7.3 million de-identified patient records, including 1.4 million with imaging data, 1.3 million with genomic information, and 260,000 with full transcriptomics profiles.

Pathos is also running its own clinical programs. In March 2025, they dosed the first patient in a Phase 1b/2a trial of pocenbrodib (a CBP/p300 inhibitor) in metastatic castration-resistant prostate cancer. They also acquired Known Medicine, which builds patient-specific 3D cell cultures and uses AI to predict drug responses prospectively — an attempt to close the loop between computational prediction and wet-lab validation.

Funding: $365 million in a Series D (May 2025), at a $1.6 billion valuation.

⭐ BPGbio

BPGbio takes a different technical angle: Bayesian causal AI, built on their NAi Interrogative Biology platform. Rather than relying on pattern recognition across large datasets, their approach aims to infer causal relationships, not just correlations, between patient characteristics and treatment response.

The platform integrates one of the largest non-governmental biobanks (over 100,000 clinically annotated patient samples) with deep multi-omic and clinical data, running on the Frontier exascale supercomputer at Oak Ridge National Labs.

The clearest demonstration of the approach in the drug rescue context comes from a multi-arm Phase Ib oncology study involving 104 patients across multiple tumor types. NAi’s models, trained on tissue and blood-derived multi-omic data, identified biological signatures predicting response to BPM31510 — and BPGbio used those insights to prioritize glioblastoma multiforme (GBM) and pancreatic cancer as the most compelling indications. The causal framing is that if the model can distinguish “patients who happened to respond” from “patients who responded because of a specific biological mechanism,” the resulting biomarkers should be more robust in prospective validation.

BPGbio is further along clinically than many AI-native biotechs. The company has completed enrollment in a Phase 2b GBM trial, with topline results expected in Q3 2026, and sought FDA guidance in late 2025 for a potential expedited regulatory path in GBM. They have multiple Phase 2 clinical trials underway — making them, by their own account, one of the first companies worldwide to advance multiple Phase 2 programs developed using causal Bayesian AI.

Share

🔹 Others in the Neighborhood

The companies above aren’t the only ones in this territory. Several others touch the thesis from adjacent angles:

🔹 NOETIK trains AI models on massive datasets of paired pathology images and spatial transcriptomics to find hidden biological subtypes among trial participants and predict which patients will respond. The pairing of imaging and spatial transcriptomics is technically ambitious and could surface subgroups invisible to genomics-only approaches.

🔹 Ignota Labs (London) takes a complementary but distinct approach. As CEO Sam Windsor has noted, sometimes the drug just isn’t good enough and Ignota focuses on fixing fundamental safety issues in the chemistry itself to give failed drugs a second chance. Patient stratification and molecular optimization are different interventions for the same problem (the 90%+ failure rate), and both are needed.

🔹 Formation Bio (New York, valued at ~$1.7 billion) acquires stalled clinical-stage drugs and uses AI to run trials more efficiently, optimizing patient recruitment, protocol design, and site management. In November 2024, they launched Muse, an AI tool for clinical trial recruitment, in partnership with OpenAI and Sanofi. Formation’s overlap with the “drug rescue” thesis is real but less precise: they’re improving trial execution, not fundamentally reanalyzing who should be in the trial. Closer to operational arbitrage than to computational patient stratification.

🔹 Origent Data Sciences uses machine learning to build patient-level predictive models for disease progression, then identifies cohorts within failed trials where treatment effects can be demonstrated. Their ForecastOne platform is specifically designed for drug rescue in neurodegenerative diseases — a space where population heterogeneity is especially pronounced.

📜 The Roivant Precedent

The idea of finding value in pharma's abandoned assets isn't new.

Roivant Sciences, founded in 2014 by Vivek Ramaswamy, was built on the thesis that the pharmaceutical industry was full of abandoned assets that failed not because of efficacy problems but because of strategic deprioritization. They licensed shelved drugs, housed them in independent subsidiaries (”Vants”), and pushed them through development.

The model worked, sometimes. The most famous case, buying GSK’s Alzheimer’s candidate intepirdine for $5 million via Axovant, failed in Phase 3. But Roivant learned and evolved. The company has since grown to a ~$20 billion market cap by pivoting to a precision focus on immunology and inflammation, and its AI story now lives in VantAI (a spinout building the Neo-1 model for molecular glue design), not in clinical failure analysis.

Roivant’s original model was a financial and operational bet, spot undervalued assets, give them focused management, and move fast. It wasn’t a computational bet on finding hidden responder subgroups in trial data per se. The companies profiled above are making a different bet: that AI can extract signal from noise in ways that traditional biostatistics couldn’t.

Roivant CEO Matt Gline has been candid about his skepticism of AI drug discovery, arguing it faces a fundamental problem by solving only one or two of the roughly 150 hard problems in preclinical development. But it was aimed at de novo drug discovery, not at the more focused application of AI for patient stratification in existing clinical data.

⚠️ Still Open Questions

We need to be honest about the risks and limitations.

The p-hacking problem is real. Post-hoc subgroup analysis is one of the oldest and most dangerous tools in clinical research. If you slice a trial population enough ways, you will find a subgroup that responded, by chance. The key question for every company in this space is: can your AI-identified subgroups be validated prospectively? Retrospective signal discovery is table stakes. Prospective confirmation is where most of these approaches will succeed or fail.

Regulatory uncertainty. The FDA has frameworks for enrichment strategies and biomarker-driven trial designs, but there’s no well-trodden regulatory path for “we reanalyzed a failed trial with AI and found a responding subgroup, now we want to run a new trial in just those patients.” Each company is navigating this largely ad hoc. Lantern’s FDA interaction on the HARMONIC trial amendments is a positive signal, but one data point doesn’t make a precedent.

Small subgroups, small markets. Patient stratification is a precision medicine play. By definition, you’re narrowing the addressable population. Some subgroups will be commercially viable (Lantern’s L858R NSCLC population of 65,000–80,000 relapsed patients per year is meaningful). Others may be too small to justify the cost of a dedicated clinical program. The economics of drug rescue only work if the subgroup is big enough, or if the development cost is low enough, to justify the investment.

Data access and quality. These approaches are only as good as the data they’re trained on. Pathos has an enormous advantage through the Tempus relationship (200+ petabytes), but most failed trials sit in corporate vaults, and the patient-level data needed for subgroup reanalysis is rarely publicly available. Companies that can’t access high-quality, multimodal, longitudinal patient data are building on thin foundations.

The “drug just isn’t good enough” problem. As Ignota’s CEO, Sam Windsor pointed out, sometimes patient stratification isn’t the answer; the molecule itself has fundamental issues. A drug with genuine safety problems or an insufficient therapeutic window won’t be rescued by finding better patients for it. The companies building these AI platforms need to be disciplined about walking away from molecules that don’t warrant rescue.

Self-reported early data. Most of the clinical results we’ve seen so far — including Lantern’s LP-300 PFS data — are company-reported, from early-stage trials, in small patient numbers. This is expected at this stage of maturity, but we shouldn’t confuse preliminary signals with validated outcomes. The real test comes in registrational trials with pre-specified subgroups and independently adjudicated endpoints.

🔭 Looking Ahead

So is “rescuing failed drugs with AI” a category now?

The capital providers certainly believe so. Between Biossil ($70M), Pathos ($365M Series D, $1.6B valuation), Formation Bio ($600M+, $1.7B valuation), and Lantern (public, running clinical trials), there’s real money behind the thesis, from investors ranging from traditional life science VCs to Founders Fund and OpenAI.

But a category needs more than capital. It needs clinical proof points. Here’s what to watch:

Near-term (2026–2027):

• Lantern’s HARMONIC trial data in the narrowed L858R NSCLC subgroup. This is one of the most concrete tests of the thesis: a drug that was broadly tested, AI-identified a specific molecular subgroup, and a redesigned trial is now running in just that population. If LP-300 produces confirmatory data, it’s a powerful proof of concept for the entire space.

• Pathos’s pocenbrodib Phase 1b/2a data in mCRPC — particularly whether their biomarker-defined subgroups show differential response.

• Biossil’s first clinical readouts from any of its ten programs.

Medium-term (up to 2028 and beyond):

• Whether any AI-identified subgroup leads to a regulatory filing. This would be the true inflection point — a drug that failed in a broad population, succeeded in an AI-defined subgroup, and got approved for that subgroup.

• The maturation of multimodal foundation models in oncology (the AstraZeneca-Tempus-Pathos collaboration). If these models can reliably predict responders across tumor types, the implications extend far beyond drug rescue.

• Whether pharma companies begin systematically reanalyzing their own shelved assets with these tools, either internally or through partnerships. The volume of failed late-stage programs sitting in corporate vaults is enormous.

The open questions:

• Can retrospective AI-driven subgroup discovery produce biomarkers robust enough for prospective trial enrichment? This is the central scientific question.

• Will regulators create clearer frameworks for AI-informed trial redesign, or will each program remain a bespoke negotiation with the FDA?

• Is there a sustainable business model here, or will drug rescue remain a niche strategy for specific asset classes? The economics depend heavily on how cheaply you can acquire failed assets, how efficiently AI can identify the right subgroup, and how large that subgroup turns out to be.

These are early days. The tools are getting dramatically more powerful, including multimodal models, massive patient datasets, causal inference frameworks, and agentic AI systems. But the clinical validation is thin, with only a handful of companies running trials. None has yet produced the definitive proof point: a failed drug, rescued by AI-driven patient stratification, approved by regulators.

We think this is a space worth watching closely, because the underlying logic is sound, the unmet need is enormous (90%+ failure rates, billions in sunk R&D), and the technical capabilities are likely approaching what the problem demands. The next 18–24 months of clinical data will tell us whether this is a genuine new paradigm or an expensive lesson in the limits of computational biology.

As always, if you’re working in this space or watching it from the inside, we’d love to hear what you’re seeing. Leave a comment!

Leave a comment

Part 1 maps the therapeutic landscape by hallmark of aging. Parts 2 and 3 will follow on biomarkers and the consumer-facing clinic ecosystem.

Part I: Therapeutic strategies that target the hallmarks of aging

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).

The industry is supported by a growing ecosystem of specialized longevity capital firms, which emerged around the idea of aging biology as an investment opportunity. Alongside Sergey Young’s Longevity Vision Fund, firms like, Apollo Health Ventures and Healthspan Capital and others are building portfolios of companies that aim to tackle the root causes and accompanying symptoms of aging. At the same time, non-profit organizations, such as Hevolution, Lifespan Research Institute and the Methuselah Foundation are offering grants and partnerships to foster breakthroughs in the longevity field, while the XPRIZE competition offers $ 101 M over 7 years for teams that develop longevity therapeutics.

However, biotechs in the longevity space face major scientific and regulatory hurdles. On the one hand, incomplete understanding of which mechanisms are causal drivers versus downstream effects of aging challenges hypothesis selection for new treatments and aging markers. On the other hand, the lack of standardized, validated biomarkers and endpoints make it hard to select the right proxies for early readouts of aging. Notably, aging itself is currently not classified as a disease indication within a regulatory approval pathway.

In part 1 of this three-article series, we explore therapeutic strategies that tackle the hallmarks of aging. In part 2, we’ll dive into aging biomarkers and biological clocks, and in part 3, we’ll analyze consumer-facing longevity companies and longevity clinics.

Therapeutic strategies for combatting aging

A number of biotechs and academic labs are tackling aging hypotheses, often organized in different iterations of the hallmarks of aging framework (Figure 2). Many of these approaches are still early-stage bets for improving life- and healthspan, and most are being tested in (age-related) proxy diseases, which offer validated endpoints and shorter development cycles.

Resetting cellular information: cellular reprogramming and rejuvenation biology

One major longevity strategy focuses on resetting cellular information through reprogramming. By reversing epigenetic drift, a process through which cells gradually lose proper gene-regulatory control, cellular reprogramming aims to shift cells from aged back to youthful phenotypes. First demonstrated by Shinya Yamanaka’s lab in 2006, reprogramming approaches are now being pursued by several well-funded biotechs, including Altos Labs, NewLimit, and RetroBiosciences.

Altos Labs launched in 2022 with around $3B in backing (some reportedly with prominent backing from Jeff Bezos and Yuri Milner) and a high-profile academic team, including stem cell researcher Juan Carlos Izpisua Belmonte, as Founding Scientist and Senior Vice President and Shinya Yamanaka, as a Senior Scientific Advisor. Altos centers on the hypothesis that partial cellular reprogramming can restore youthful function without erasing cell identity, but it has remained largely stealthy about its specific programs. In May 2025, Altos acquired senescence-focused Dorian Therapeutics.

A first example of clinical testing in the area of reprogramming is Harvard scientist David Sinclair’s Life Biosciences, which has cleared the IND for ER-100, a partial epigenetic reprogramming program for optic neuropathies and closed an $80 million Series D financing in April of 2026.

In pursuit of more effective reprogramming, several companies are leaning on artificial intelligence. NewLimit, Shift Bioscience and clock.bio, for example, use AI-driven approaches to better understand (epi)genetic programs of aging and improve reprogramming strategies. NewLimit focuses on reprogramming cells to a younger state using AI-guided discovery, with a near-term emphasis on liver-directed mRNA reprogramming. NewLimit, co-founded by Coinbase CEO Brian Armstrong, raised a $130M Series B in May 2025 to push lead programs toward the clinic. Shift Bioscience, a Cambridge (UK)-based company, uses generative AI and virtual cell approaches to identify gene programs in single-cell analyses that can rejuvenate cells. In October 2024, it raised a $16M seed round to scale the platform and advance targets toward drug development.

San Francisco–based longevity biotech Retro Biosciences (backed by OpenAI’s Sam Altman, who led a$180M at seed and reportedly participated in the company’s ~$1B Series A in 2025) has positioned AI as part of its strategy for designing rejuvenation factors. Retro has begun moving its first candidates toward human testing, with a program centered on autophagy (another hallmark of aging) having already entered phase 1 in late 2025.

Notably, some companies such as Turn.Bio and Pelage Pharmaceuticals are pursuing dermatology as a near-term, measurable indication for rejuvenation. Pelage Pharmaceuticals aims to reactivate or restore follicle function to counteract hair loss, and announced a $120M Series B in October 2025 to advance its lead program through clinical development.

Removing or neutralizing harmful aged cells: senotherapeutics

Senescence is the process by which cells undergo an often irreversible cell-cycle arrest. While beneficial in certain contexts, senescent or “zombie” cells can accumulate and drive tissue dysfunction through inflammatory signaling via the SASP (senescence-associated secretory phenotype), contributing to chronic inflammation (inflammaging) and age-related tissue decline. So-called senotherapeutics try to remove these cells (senolytics) or blunt their harmful signaling (senomorphics).

AI–techbio pioneer Insilico Medicine’s TNIK inhibitor, ISM001-055, which is currently being tested in idiopathic pulmonary fibrosis (IPF) has shown senomorphic potential in a recent study. ISM001-055 (rentosertib) is one of the first examples of an AI-designed small molecule advancing in human trials, with positive efficacy and safety readouts from a randomized Phase 2a study. By potentially attenuating cellular senescence, the drug could help suppress multiple aging-related processes, suggesting broader applications as a senomorphic therapy in age-related diseases beyond fibrosis.

Rubedo Life Sciences aims to tackle senescent cells by selectively inducing regulated cell death via ferroptosis. The company, which raised $40M in a 2024 Series A, has an AI-driven drug discovery platform that employs single-cell RNA sequencing to drive small-molecule design for senescence-specific targets. Rubedo’s lead program, RLS-1496, a first-in-class selective GPX4 modulator, will reportedly commence clinical testing in psoriasis, atopic dermatitis, and skin aging. SENISCA’s senotherapeutic platform builds on the hypothesis that senescent cells show dysregulated patterns of RNA splicing. The company that spun out of the University of Exeter (UK) announced £3.7M in seed financing to advance its splicing-modulation platform and early pipeline.

Another company that operated in the senolytic space was Unity Biotechnology. However, despite initial hints of efficacy for its senolytic program, UBX1325 (foselutoclax) for diabetic macular edema, a phase 2 trial did not meet its primary endpoint and the company is no longer operating, having finalized liquidation in September 2025.

Reframing the metabolic hub: targeting mTOR and nutrient sensing pathways

Genetic analyses, lifestyle intervention, and drug treatment experiments in animals have consistently pointed toward metabolic pathways as key players in longevity, sitting at a central hub between metabolic control, growth regulation, inflammation, and stress pathways.

mTOR inhibitors like rapamycin were among the first compounds recognized to improve longevity in mice and are widely used in the longevity biohacking community. Rapamycin’s impact on aging is currently being tested in a double-blind, placebo-controlled, multicenter trial in healthy, middle-aged companion dogs in scope of the Dog Aging Project. Companion dogs are seen as good models for human longevity due to metabolic and lifestyle similarities.

Companies like Cambrian and Aeovian Pharmaceuticals are testing mTORC1 inhibitors in age-related proxy diseases. Aeovian raised $55M in December 2025 to advance its lead, a selective, CNS-penetrant mTORC1 inhibitor in Tuberous Sclerosis Complex (TSC)-related epilepsy. Similarly, Cambrian recently announced up to $30.8 M in funding from Advanced Research Projects Agency for Health (ARPA-H) to develop its selective mTORC1 inhibitors, in line with the agency’s PROSPR program.

In addition to mTOR inhibitors, diabetes drugs like metformin and GLP-1 agonists have been linked to longevity in various contexts. While the large-scale trial of the diabetes drug metformin, TAME (Targeting Aging with Metformin), as a longevity-promoting factor is still awaiting recruitment, a number of studies in patients with metabolic syndrome or obesity have shown benefits of GLP-1s on metabolic comorbidities, which are often associated with age-related diseases, including reductions in major adverse cardiovascular events and kidney outcomes. However, a beneficial effect of GLP-1s on neurodegenerative diseases could not be confirmed in a Phase 3 trial for slowing Alzheimer’s progression. As pointed out in a recent commentary in Nature Health, rigorous testing of GLP-1 receptor agonists as antiaging drugs will be required to show if they live up to their potential.

Cardiometabolic health is recognized as a specific risk factor in the metabolic pathways of aging. BioAge Labs employs its platform linking longitudinal multi-omics with healthspan trajectories (with aging cohorts followed for up to 50 years) to find druggable targets that are linked to metabolic aging. The company recently refocused on its early-stage, broader cardiometabolic aging pipeline including NLRP3 inhibition, after its lead obesity asset (azelaprag) showed liver toxicity signals in a Phase 2 study months after the company’s $238M IPO in late 2024.

Other biotechs in the cardiometabolic space are specifically targeting atherosclerosis as an age-related disease. For example, Cyclarity Therapeutics is built around the hypothesis that clearing arterial cholesterol toxins can reverse atherosclerotic disease biology. In Jan 2025, it announced the first tranche of a Series A and a first-in-human study for UDP-003, a candidate aimed at addressing cardiovascular risk via this mechanism. Repair Biotechnologies is developing first-in-class therapies that could reduce atherosclerotic plaque by selectively clearing excess intracellular cholesterol. The company is developing its mRNA therapy REP-0003 for Homozygous Familial Hypercholesterolemia, a rare condition of accelerated atherosclerosis, and received FDA Orphan Drug Designation in May 2025.

Restoring cellular bioenergy: mitochondrial function

Mitochondrial dysfunction is an aging hallmark and interacts with multiple other hallmarks. Bioenergetics restoration approaches target mitochondrial function and energy production to improve resilience in age-related disorders. Pretzel Therapeutics targets diseases where mitochondrial dysfunction is causal, based on the hypothesis that restoring mitochondrial function and mtDNA maintenance can treat rare mitochondrial diseases and may translate into broader age-associated conditions. In April 2025, the biotech initiated Phase 1 recruitment for PX578, a first-in-class approach targeting mitochondrial DNA polymerase. Similarly, MitoRx Therapeutics’s mitochondrial metabolic modulation platform has the potential to address obesity and other key cardiometabolic diseases. An interesting approach toward improved mitochondrial function is being taken by California-based startup mitrix. The company aims to grow mitochondria in large quantities in a bioreactor, with the goal of targeted, organ-specific replacement of functional mitochondria, with initial exploratory human testing started in 2025, including a 90-year old physics professor emeritus from the University of Washington.

Reinstating immune function: Targeting age-related immune dysfunction and stem cell exhaustion

Aging is often accompanied by innate and adaptive immune dysfunction (a process termed immunosenescence), which leads to higher infection susceptibility and lower vaccine efficacy. Moreover, systemic, low-grade inflammation levels (inflammaging) constitutes a major risk factor for aging-related diseases, linked to senescence and metabolic aging hallmarks.

Immunis, which announced a $25M Series A-1 in January 2025, is developing therapies to tackle failure of immune-regenerative signaling in aging-related disease. Philadelphia-based biotech Tolerance Bio aims to develop therapies for immune-mediated diseases, by exploiting the function of the thymus as a key regulator of immune tolerance. The company, which launched with $17.2M in seed financing in October 2024 plans to develop both cell-based therapeutics and RNA-based therapeutics for age-related disease in collaboration with Zipcode Bio. Similarly, TECregen, Thymmune Therapeutics and Intervene Immune aim to explore thymus biology to counteract age-related immune decline. Swiss biotech TECregen, which launched with $12.6 M (CHF 10 M) seed financing set out to develop thymopoietic drugs that restore immune balance, and address immune decline caused by aging.

A number of companies such as HERVolution Therapeutics and Transposon Therapeutics target mobile elements within the dark genome that can become reactivated with age or disease, from remnants of ancient retroviral infections to autonomous retrotransposons. Copenhagen-based HERVolution raised $11.7M in Series A to advance immunotherapies targeting human endogenous retroviruses (HERVs) in December 2024, while Transposon was awarded $22M by the ARPA-H for its lead TPN-101 to be developed under the PROSPR program.

Other companies, such as ImmuneAge and MoglingBio tackle immune function upstream by employing strategies for rejuvenating hematopoietic stem cells. German startup Mogling Bio is developing pharmacological rejuvenation strategies for restoring the function of “exhausted” hematopoietic stem cells. The company targets the small GTPase Cdc42, which becomes hyperactive with age, following the hypothesis that normalizing age-associated Cdc42 overactivation could help recover stem cell polarity and function. Downstream, this strategy could improve blood and immune cell quality, positioning stem cell rejuvenation as a potential intervention lever for immunosenescence.

Taming chronic inflammation: Targeting inflammaging as a driver of age-related disease

Other companies, such as Utah-based startup Halia Therapeutics, which raised $30M in a Series C in January 2024 tackle inflammaging, by targeting key players in inflammatory cascades. The company’s lead asset, ofirnoflast (HT-6184), is a NEK7-based modulator of NLRP3-mediated inflammation. The compound received Orphan Drug Designation from the FDA for myelodysplastic syndromes (MDS) in October 2025, with positive Phase 2a data presented at ASH 2025.

Alphabet’s Calico also just made a big bet on targeting inflammation in age-related disease. To complement its portfolio, which includes a number of bets on age-related diseases and an age-related disease alliance (worth up to $3.5bn) with AbbVie, Calico inked a $596M deal (including a $25 million upfront) with Chinese antibody drug company Mabwell for its IL-11 inhibitor. The anti-inflammatory cytokine IL-11, which acts on the ERK–AMPK–mTORC1 axis, is upregulated in aging cells and its inhibition was shown to extend health- and life spans in animal models.

Another company that aims to tackle the intersection of inflammation and fibrosis is AI- and longevity-focused biotech Juvenescence. The company’s plasminogen activator inhibitor-1 (PAI-1), recently completed a phase 1 safety study and is supposed to enter a Phase 2 proof-of-concept trial planned for in patients with metabolic and fibrotic disease.

The future of longevity drug development

While a number of companies are centered around specific aging mechanisms, others such as Calico, Retro Biosciences, and Insilico Medicine are diversifying across different aging mechanisms and manifestations. Other companies focus on a specific mechanism that is potentially applicable across different aging-related diseases. LinkGevity is creating anti-necrotic drugs using AI-driven approaches. Elevian focuses on restoring youthful regenerative capacity in humans by exploiting the biology of GDF11, a circulating blood factor shown to increase neovascularization and neurogenesis in mouse brains, initially focusing development on stroke with potential for expansion into cardiometabolic and inflammatory disease.

Across varying hypotheses and modalities, the longevity biotech field appears to be converging on a pragmatic strategy that bridges the realities of drug development with the ambitions of increased life- and healthspans: prove a mechanism in a recognized (aging-related) disease, measure something that moves earlier than mortality, explore long-term potential for aging. Interestingly, as a near-term proxy for studying interventions for human aging, some biotechs like Genflow Biosciences, Loyal and Rejuvenate Bio are testing therapeutics in companion dogs as a potential “first approval” pathway for an explicit longevity indication Loyal (which gathered $100M Series C funding in February 2026) is testing its LOY-002 for healthier lifespan extension in senior dogs and announced FDA CVM acceptance of a key safety package for LOY-002 in January 2026. Rejuvenate Bio develops gene therapies in animal health, including a gene therapy for canine osteoarthritis. with a long-term vision that veterinary translation can de-risk platforms relevant to human aging biology.

Alongside drug-based geroscience interventions, a growing set of academic groups and companies such as Renewal Bio, BE Therapeutics, United Therapeutics, eGenesis, New Jersey-based Celularity and Bayer’s BluerockTherapeutics are pursuing regenerative and replacement approaches, based on the idea that some age-related disease can be addressed by restoring lost cells and tissue function (via stem cells or engineered tissues) or by replacing failing organs, including xenotransplantation approaches. Notably, regenerative medicine biotech Longeveron just reported improvement in age-related frailty from a phase IIb study of their allogenic mesenchymal stem cells therapy laromestrocel. In their publication in Cell Stem Cell in March 2026, the authors indicate a clinically meaningful, dose- and time-dependent increase in the 6-min walk test for individuals with frailty treated with laromestrocel.

In light of the growing burden of chronic disease and the socioeconomic payoffs of increasing human healthspan, investment in longevity should be a priority for healthcare systems worldwide. Indeed, initiatives like ARPA-H’s PROSPR program that recently dedicated over $144 M in funding to interventions intended to prolong resilience with age , and the European Innovation Council naming “Biotechnology for Healthy Ageing” among its 2026 Pathfinder Challenges, hint at increased public recognition of healthspan investments. At the same time, the public funding for aging-related research remains on shaky ground. The FY2026 President’s Budget request for the U.S. National Institute on Aging would reduce the institute’s funding by 40.5% compared with the FY2025 level. To truly put longevity on the agenda of healthcare systems, public and private investments that foster longevity therapeutics as well as preventative strategies will be key.

Subscribe now

References

1. https://ourworldindata.org/data-insights/global-average-life-expectancy-has-more-than-doubled-since-1900

2. https://www.nature.com/articles/s43587-024-00702-3

3. https://pubmed.ncbi.nlm.nih.gov/23746838/

4. https://pubmed.ncbi.nlm.nih.gov/39418098/

5. https://www.nature.com/articles/s41467-023-39786-7

6. https://pubmed.ncbi.nlm.nih.gov/16904174/

7. https://pubmed.ncbi.nlm.nih.gov/37118377/

8. https://pubmed.ncbi.nlm.nih.gov/40563501/

9. https://www.nature.com/articles/s41591-025-03743-2

10. https://www.nature.com/articles/nature08221

11. https://www.nature.com/articles/s41587-025-02932-1

12. https://www.nature.com/articles/s44360-026-00109-x

13. https://pubmed.ncbi.nlm.nih.gov/34425083/

14. https://pubmed.ncbi.nlm.nih.gov/37179335/

15. https://www.nature.com/articles/s43587-025-00938-7

16. https://pmc.ncbi.nlm.nih.gov/articles/PMC10830426/

17. https://pmc.ncbi.nlm.nih.gov/articles/PMC12259695/

18. https://www.nature.com/articles/s41586-024-07701-9#Sec4

Read also

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.

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?

Subscribe now

🤖 Why agents, and why now?

A historical detour. The term “agent” gets used loosely enough in AI marketing that it might be worth tracing from its original meaning. The ideas behind it were actually tested long before today’s language model AI existed. The fundamental idea behind an agent is a feedback loop where a system perceives its environment, observes changes and adjusts in response.

Norbert Wiener and W. Ross Ashby worked on this in the 1940s, doing cybernetics. Their framework kept coming back to one idea that effective control depends more on the quality of the feedback loop than on the sophistication of the controller. Even a simple device like a thermostat qualifies: it doesn’t need to be smart, it needs a clean reading and a reliable switch.

For roughly three decades after that, the dominant AI paradigm assumed the opposite: that intelligence requires building an internal symbolic model of the world and then reasoning over it. Sense the environment, build a representation, plan against it, act. This was sometimes called GOFAI (“Good Old-Fashioned AI,” John Haugeland in 1985), and it produced systems that could play chess and prove theorems but later couldn’t walk across a room without tripping.

By the late 1980s, Rodney Brooks at MIT was building robots that dispensed with internal world models entirely. These had layered behaviours (avoid obstacle, follow wall, seek light) that composed into complex action without any central planner.

His argument against the symbolic AI mainstream was that intelligence doesn’t live inside the agent. It comes from the agent’s relationship with the environment. In “Elephants Don’t Play Chess” (1990), he wrote:

The world is its own best model—always exactly up to date and complete in every detail.

A simple agent in a well-structured environment beats a complex one in a poorly structured one.

Through the 1990s, multi-agent systems became a formal subfield concerned with how to coordinate many autonomous software agents, each with limited capabilities, so that useful collective behaviour emerges. The Belief-Desire-Intention models taken from philosophy and applied to software gave individual agents beliefs about the world, desires they wanted to achieve, and intentions they committed to. Swarm algorithms showed that coordination could arise without any agent understanding the whole and air traffic simulations demonstrated the approach at scale.

When returning our attention to the modern day version of LLM-based AI, let’s remind ourselves that, at its core, a large language model predicts text. Fittingly enough, it got good at this through human feedback during training.

And here is where the circle closes. We spent decades scaling the internal capability of AI systems, built the largest, most capable text-prediction machines in history, and the moment we try to make them do things in the world, act on observations, use tools, adjust to what happens next—the oldest insight in the field loops right back on us. To make an LLM good at acting (and, perhaps, closer to intelligence), we are back to feedback loops.

💭 Agents today

The current, fashionable incarnation of this idea is an LLM with access to tools.

Instead of a chatbot that answers questions, an agentic system acts. It breaks a goal into subtasks, calls external tools at each step (e.g. databases, APIs, code execution environments, other agents), and, ideally, carries context across the chain without losing the thread.

In biotech and pharma, this can map onto things like target identification, literature mining, data extraction, safety profiling, trial design, and regulatory documentation, all run through separate teams, separate tools, and separate institutional memories. An agentic system can serve as connective tissue across these, operating at a speed and parallelism higher than any single team.

Why now? A few reasons behind the current momentum:

1. Context windows grew large enough that models can now hold meaningful complexity in a single reasoning chain.

2. Tool-use capabilities matured: previously, every agent needed custom connectors to every data source, which has now moved closer to plug-and-play.

3. Inference cost dropped. According to Epoch AI, the price of achieving a given level of model performance has been falling by 10x to 900x per year, depending on the benchmark. GPT-4-level performance that cost $20 per million tokens in late 2022 now runs at roughly $0.40.

4. Open-source agent ecosystem exploded. From orchestration frameworks like LangChain, CrewAI, and AutoGen to full personal-agent runtimes like OpenClaw, the barrier to building an agentic system dropped considerably. In many cases, there’s no need to wire everything from scratch.

As a timely demonstration from pure machine learning recesses, the other day, Andrej Karpathy (former head of AI at Tesla and one of the original OpenAI researchers) open-sourced a minimal setup where an AI agent modifies code, runs a five-minute ML experiment, checks if the result improved, keeps or discards, and loops. Running on a single GPU node (that’s still a lot of compute), he left it iterating for two days and came back to ~20 improvements that all held up.

Karpathy called it ‘wild’ as he’d spent two decades doing exactly this kind of iterative neural-net tuning manually, and his very first naive attempt with the agent already beat what he considered a well-tuned project. His read on where it leads is that every frontier lab will do this, spinning up agent swarms that collaborate to tune models at increasing scale and “...humans (optionally) contribute on the edges.”

⏪ Last we checked

When we surveyed the landscape of AI agents in biotech last spring, the honest summary was this:

• Early-stage, fragile, mostly academic.

• A handful of systems had demonstrated interesting capabilities like TxAgent (Harvard), BioDiscoveryAgent (Stanford), SpatialAgent (Genentech), and Causaly’s knowledge graph agents.

• None were in production, tool chains were brittle, costs were steep, and there was no regulatory framework for any of it.

The field was caught between two realities of impressive demos on one side, and on the other, as always, the irreducible complexity of biology. But things changed quite fast.

⏳️ What changed?

The most visible change is what’s happening with the hardware and the scale of investment, starting with infrastructure.

🔹 Eli Lilly went live with LillyPod in late February with over 1,000 GPUs delivering 9,000+ petaflops. That followed a $1 billion co-innovation lab with NVIDIA announced in January. Through Lilly’s TuneLab platform (with access to models trained on ~$1B Worth of proprietary drug discovery data), select models will be available to biotech partners via federated learning where partners train on Lilly’s models with their own data, without transferring it.

🔹 Going even higher on compute, Roche just announced the deployment of over 3,500 GPUs across the U.S. and Europe, the largest announced GPU footprint in pharma. Genentech’s Aviv Regev framed it around Roche’s “Lab-in-the-Loop” strategy, which she said they have pursued for more than five years. An NVIDIA pre-briefing offered some early concrete numbers: nearly 90% of eligible small molecule programs at Genentech now integrate AI, and at least one molecule was designed measurably faster.

🔹 Extending into the instrument layer, Thermo Fisher and NVIDIA announced a strategic collaboration at start of this year to develop AI-native laboratory workflows and instrumentation.

But the hardware is ahead of the results. Lilly’s Diogo Rau told CNBC last October that AI-assisted benefits would likely materialize around 2030. At the LillyPod inauguration, he was cautious: “The hype is actually a serious threat to the research itself. Because if the hype becomes the story, then we’re all going to be disappointed.” The infrastructure is there, but the public record still contains far more detail on compute scale than on named downstream outputs. Roche has described at least one molecule whose redesign was accelerated.

That’s the infrastructure investment. What about actual use?

Pharma appears to see the first value of agentic AI in fixing data and workflow mess, not in autonomous discovery:

1. In the Owkin/STAT survey, 37% called implementation “very important,” but only 3% said it was the number one priority. More importantly, respondents put data challenges first at 41.6%, ahead of early discovery at 28.7%.

2. Deloitte‘s September 2025 survey of 100 U.S. healthcare technology executives found a similar pattern from the budget side: 61% were already building agentic AI initiatives or had secured funding, and 85% planned to increase investment over the next two to three years.

3. The Microsoft-NEJM AI report, focused on health systems, found actual deployment even thinner—just 3% of 30 surveyed organizations, with 43% still in pilots.

The numbers above show that deployment is thin, but there are already a few visible examples.

➕ AstraZeneca

AstraZeneca published one of the first honest accounts of putting an agentic system into a real pharma pipeline. Their paper describes ChatInvent, a conversational interface for drug discovery that evolved from a single-agent proof of concept into a multi-agent architecture. The paper is notable for what it says about how things break: every LLM upgrade required at least a week of prompt re-tuning and could change agent behavior unpredictably. The supervisor agent would silently mangle inputs, sub-agents would sometimes refuse tasks they were perfectly capable of handling. The multi-agent system was faster and cheaper than the single-agent one, although it also introduced more errors.

➕ IQVIA

At GTC 2026, IQVIA launched a unified agentic platform built with NVIDIA that bundles over 150 specialized agents for clinical, commercial, and real-world evidence workflows. The collaboration with NVIDIA dates back over a year; one of the earlier agents, a clinical data review orchestrator first shown at GTC Paris in mid-2025, uses automated checks and sub-agents to catch data issues early, cutting the review cycle from seven weeks to two. The initial release covers trial start-up, target identification, data review, market landscaping, and field sales preparation, with more agents expected in Q4.

➕ Daiichi Sankyo

Daiichi Sankyo has been using AI built with BCG to personalize responses to patient and HCP queries inside its Veeva-based systems across Europe and Canada. It's a more commercial deployment than AstraZeneca's experiment or IQVIA's agents, with content generation and protocol writing on the roadmap for 2026.

➕ Visions, and Others

There’s recent Insilico Medicine and Eli Lilly paper worth flagging that belongs in a more of a ‘vision’ category for now. Their February “From Prompt to Drug” paper describes a fully autonomous pipeline where a central reasoning controller coordinates specialized AI agents across target discovery, generative chemistry, automated synthesis, and clinical planning in a single closed-loop workflow. A scientist types a prompt; the system orchestrates the rest. The authors acknowledge the end-to-end vision “may seem far beyond what is possible today,” and argue the individual building blocks already work at a smaller scale.

➕ And then, there’s the FDA

The FDA also deployed agentic AI capabilities for its own staff in December 2025, though the rollout has been bumpy. The agency’s earlier gen AI tool, Elsa, was seen fabricating nonexistent studies and misrepresenting research, with employees describing it as unreliable for anything beyond meeting notes. All of this against the backdrop of over 1,000 staff cut from the drug review center and multiple missed approval deadlines. The regulator is experimenting with the same tools it will eventually have to regulate, and running into the same problems.

📝 A recent review in Drug Discovery Today suggests agentic AI is already valuable in drug discovery, just not where most of the headlines are pointing.

The paper covers eight case studies from companies including Potato, Plex Research, and Coincidence Labs, several with specific quantitative benchmarks. The gains that hold up are in the operational middle of discovery: literature synthesis, protocol generation, assay design. Potato’s Tater agent took the design cycle for a qPCR assay from one-to-four months down to under two hours, although empirical validation was still needed.

The agent architectures in these systems mirror how discovery teams already work— supervisor delegates to specialists, shared context, iterative refinement—just without the multiweek meeting cadence. The authors note that current benchmarks capture whether an agent got the right answer but not whether the reasoning behind it was sound, and that early-stage results like cell-level inhibition don’t guarantee downstream translation.

One gap: every case study reports time savings, none report what the infrastructure costs to build and run. But the broader view is that the real value right now is compression of the coordination overhead between steps that already work on their own, not autonomous science.

📌 What does this add up to?

The infrastructure is real, the investment is committed, and a handful of systems are actually running in production workflows. But the gap between the hardware announcements and the named scientific outputs is still wide, and the honest accounts show that multi-agent systems in messy real-world pipelines are notoriously fragile. The most grounded read right now is that agentic AI in biopharma is just right on cusp of leaving the “interesting demos” phase but well short of the “reliable infrastructure” phase.

🧪 Agents doing science

A single LLM asked to check its own work tends to agree with itself. In one medical study, some frontier models (in 2025) complied with illogical requests up to 100% of the time. Most multi-agent systems built over the past year have had to engineer around this. The solutions vary, but they all converge on the same idea of creating friction.

• Google‘s AI co-scientist uses what it calls a “generate, debate, and evolve” framework where agents propose hypotheses, other agents critique them, and the survivors get refined through iteration. In a collaboration with Stanford, two of three co-scientist-recommended drugs for liver fibrosis showed “significant anti-fibrotic activity” in human liver organoids.

• DeepMind‘s Aletheia has a generator-verifier-reviser loop where agents produce solutions, check them for flaws, and correct or discard faulty reasoning.

• Stanford‘s Virtual Biotech (the one we opened with) assigns a dedicated reviewer agent that evaluates outputs and pushes back. The same lab is also trying out the opposite approach with a generalist single Biomni agent that skips the team structure entirely, composing its own workflows across 25 biomedical subfields.

• FutureHouse, an Eric Schmidt-backed nonprofit in San Francisco building what it calls an “AI Scientist,” went wider: a single up-to-12-hour run executes up to 42,000 lines of code across 166 data-analysis agent rollouts and reads roughly 1,500 papers across 36 literature-review agent rollouts, coordinated through a structured world model. It reported seven discoveries, four of them “novel,” and three that independently reproduced unpublished human findings.

Even though friction helps, it doesn’t solve the problem entirely. Just from the systems above: Kosmos’s reports were rated about 79% accurate by independent scientists, but FutureHouse’s own team notes the system often chases statistically significant but scientifically irrelevant findings. Aletheia ran through all 700 open Erdős problems in a week, its verifier flagged 212 as potentially correct, human experts confirmed 63 as technically valid, but only 4 resolved genuinely open questions. Sakana AI, about a year ago, produced what it called the first fully AI-generated paper to pass peer review, but the caveat here is that it was a workshop submission, humans selected which generated papers to submit, and the paper was withdrawn.

Model providers are, of course, ambitious and optimistic.

🔹 OpenAI told MIT Technology Review that building a fully automated AI researcher is now its explicit priority with an “autonomous research intern” by September, a full multi-agent system by 2028. Its chief scientist Jakub Pachocki described a future where a “whole research lab” exists inside a data center. Doug Downey at the Allen Institute for AI calls the prospect “exciting” but cautions that multi-step scientific work compounds error, and chaining tasks makes success less likely across the whole sequence.

🔹 Anthropic is focusing less on a standalone autonomous researcher and more on embedding Claude into existing scientific workflows through partners like HHMI’s Janelia campus and the Allen Institute, while extending into research infrastructure and biopharma R&D through its Claude for Life Sciences rollout.

So while some push toward replacing the process entirely, others look to situate the tools inside an already existing human/institutional process.

Continuing with limitations—when agents are all instantiations of the same underlying model (or similar models), their “disagreement” is bounded by shared priors, shared training data, and shared failure modes. They’re unlikely to catch each other’s systematic blind spots, and only catch surface-level inconsistencies—a so-called ‘tyranny of the majority’ where homogeneous agents converge on shared errors unwittingly. None of them can encounter surprise.

A real experiment can falsify a hypothesis in a way no agent in the loop anticipated. Multi-agent critique can’t replicate that. Conveniently, a real science lab provides that ‘for free’, as some friction with reality is inherent to it.

♻️ The lab-in-the-loop

Let’s recall Rodney Brooks when he wrote that “the world is its own best model.” A wet lab isn’t quite the world, but it’s a lot closer to it than AI agents debating themselves virtually. Ross King‘s Robot Scientist ‘Adam’ was already doing something close to this in 2009 by formulating hypotheses, running physical experiments, interpreting results, and confirming novel gene functions in yeast without a human in the loop.

Lila Sciences’ CTO Andrew Beam frames a certain bottleneck we’ve now reached: AI advanced fastest in domains where results are “easy to verify,” like mathematics, where proofs can be checked mechanically. Science doesn’t offer that shortcut, so verification means running an experiment. Beam posits that boosting the throughput of experiments describing the physical world will provide the critical data stream for the next generation of AI models.

The approach works best when three conditions align: a large combinatorial space, automatable chemistry, and a fast quantitative readout.

A decent example of this is LUMI-lab—a self-driving platform for ionizable lipid discovery recently published in Cell. A foundation model pretrained on 28 million molecular structures proposes candidates, robots synthesize and test them, and the results feed back in. One design-make-test-learn cycle every 39 hours. Over ten rounds, the system evaluated over 1,700 lipids, and by round ten more than half exceeded the transfection efficiency of MC3, a clinical-grade benchmark. The top compound achieved 20.3% gene editing in mouse lung epithelial cells via inhalation, reported as a new bar for inhaled CRISPR delivery. Humans still handle hardware errors and interpret edge cases, but the experimental loop itself runs unattended.

Not every lab-in-the-loop system aims for full autonomy: Le Cong’s (Stanford) and Mengdi Wang‘s (Princeton) LabOS keeps the researcher in the loop and augments them instead—AI agents connected via smart glasses and robots read experimental context and assist in real time, extending their earlier CRISPR-GPT work into the physical lab.

Ginkgo Bioworks and OpenAI report they connected GPT-5 to Ginkgo’s cloud laboratory and optimized cell-free protein synthesis across six iterative rounds over six months, testing over 36,000 reaction compositions. They say the system reduced production cost by 40% relative to prior benchmarks. One important caveat is that the results were demonstrated on a single protein (sfGFP), and when tested on twelve additional proteins, only half were even detectable. Ginkgo is already selling the AI-improved reagent mix commercially.

Roche’s “Lab-in-the-Loop” strategy, Lilly’s integration of agentic AI with robotic biomanufacturing, and Lila Sciences’ autonomous labs are all aimed at this kind of continuous computation-experiment cycle. The barrier to entry is also dropping because cloud lab platforms like Strateos and Emerald Cloud Lab let smaller teams plug into robotic infrastructure without building their own.

📌 Reality checkpoint

The gap between what these systems can do and what they’re being described as doing is still wide. The idea of lab-in-the-loop is more trustworthy than pure virtual debate, but it only works when the problem is shaped right. Most biology either isn’t shaped right or is hard to shape. The more durable near-term bet is probably the less glamorous one: embedding these tools inside existing scientific institutions rather than replacing the process completely, accepting that human judgment stays load-bearing for now, and letting the autonomy expand incrementally as reliability earns it.

⛓️‍💥 What doesn’t work

Now, back to limitations.

• Reliability. A February 2026 paper from Princeton and Cornell evaluated 14 agentic models and found that nearly two years of rapid capability gains have produced only modest improvements in reliability. Agents that can solve a task often fail on repeated attempts under identical conditions, with outcome consistency scores ranging from 30% to 75%. All three major providers clustered together. Scaling up didn’t uniformly help: larger models improved calibration and robustness but actually hurt consistency, showing more run-to-run variability. An agent that passes a benchmark may behave differently each time you run it on the same input.

• Architectural narrowness. A systematic evaluation of six drug discovery frameworks (Wijaya, Feb 2026) found all six locked into the same pattern: LLM reasons over text, calls APIs. That works for literature review and SMILES-based molecular design. It breaks when we need what drug discovery actually requires: model training, reinforcement learning, simulation, in vivo data integration, multi-objective optimization. The bottleneck isn’t really model knowledge (frontier LLMs reason about peptides competently) but the fact that no framework exposes those capabilities. There’s also a resource assumption baked in misaligned with small biotech realities: all six frameworks assume large-pharma data volumes, cluster-scale compute, and specialized teams.

Having several LLMs critique each other’s reasoning is the most discussed mitigation. The evidence is growing, and it’s mixed.

• Estornell and Liu (NeurIPS 2024) formalized the core problem as “tyranny of the majority”: when most agents share a misconception, minority agents conform rather than push back. The echo chamber follows from models sharing training data, priors, and failure modes.

• Wynn and Satija (2025) went further, showing that debate can actively degrade performance—models shifted from correct to incorrect answers by favoring agreement over challenging flawed reasoning, even when the stronger model outnumbered weaker ones.

• Wu et al. (2025) confirmed the pattern from a different angle: in controlled experiments, intrinsic reasoning strength and group diversity drove debate success, while structural tweaks — depth, turn order, confidence reporting — did little. You cannot scaffold your way past weak reasoning.

• Mixed-vendor teams help. Yuan et al. (Feb 2026) showed that assembling agents from different model families consistently outperformed single-vendor teams in clinical diagnosis, catching blind spots that homogeneous teams reinforced. But this is diversifying error profiles, not eliminating error.

📑 The echo chamber has an upstream version—Marinka Zitnik, associate professor of biomedical informatics at Harvard, noted in a recent Fay Lin’s GEN feature that 95% of all life sciences publications focus on roughly 5,000 of the most well-studied human genes. An agent trained on that literature will generate hypotheses that cluster around the same targets, not because the model lacks reasoning ability but because the knowledge base is lopsided.

Diversifying the model vendor doesn’t fix a skewed training signal. What does, at least partially, is tying agents to data modalities the literature underrepresents—single-cell sequencing, molecular structures, longitudinal clinical trajectories—which is another way of saying: back to the lab.

And there’s more:

• Tool fragility. In AstraZeneca’s case, every LLM upgrade required at least a week of prompt re-tuning, supervisor agents mangled inputs, sub-agents refused tasks they could handle, and multi-agent setups introduced more errors than single-agent ones. A single unexpected API response crashes a multi-step chain. Recovery is ad hoc.

• Error compounding. An agent that’s 95% accurate per step drops below 60% over a ten-step chain.

• Preclinical speed is not total speed. AI compresses early discovery timelines. It does not compress clinical trials, patient enrollment, regulatory review, or biology itself.

Presently, agents are getting better at talking about science faster than they’re getting better at doing it.

🔒 What can go wrong

While ‘what doesn’t work’ is about epistemological and performance failures like reliability, narrowness, hallucination, echo chambers, regulatory gaps (failures in a benign environment)—there’s also ‘what can go wrong’ adversarial failure—what happens when someone is actively trying to break or exploit the agent?

Agents with access to clinical data, lab automation systems, and regulatory documents present an attack surface that we are only starting to reckon with. Cisco’s State of AI Security 2026 report found that only 29% of organizations felt prepared to secure agentic deployments.

Anthropic reported that in mid-September 2025 it detected what it described as the first documented large-scale cyberattack executed without substantial human intervention, targeting roughly thirty entities across sectors including finance and chemical manufacturing through manipulated Claude Code.

In pharma, where a compromised agent could alter experimental protocols, misroute regulatory filings, or leak proprietary compound data, the consequences are sector-specific and hard to bound. The now-(in)famous “move fast and break things” motto that somewhat works in consumer software carries a different risk profile here.

🔭 Looking ahead

The tools evolved, deployments are growing (if thin), and many are building their own or adding on agents. The AI infrastructure commitments are serious, and so is the gap between what’s been announced and what’s been shown to work. Somewhere between a gold rush and a correction there are a few things to look out for:

• Regulation. This January, the FDA and EMA jointly identified ten principles for good AI practice across the medicines lifecycle, spanning work from early research through post-market activities. The EU AI Act’s high-risk provisions are now coming into force, with healthcare AI in scope. But neither touches agentic AI specifically. Autonomous agents that plan, chain tools, and act across multi-step workflows present a different challenge that the current frameworks haven’t caught up to.

• First regulatory submission with agentic contributions. At some point, someone will file an IND where agents meaningfully contributed to the evidence package, target selection, data analysis, or safety profiling. When a regulator has to evaluate that and decide what counts as adequate documentation of what the agent did and why, there will be a conversation around audit and accountability.

• Interoperability standards. Seed-funded by Genentech, The Pistoia Alliance is building agent-to-agent communication protocols for life sciences. Although most companies aren’t yet at the stage where cross-vendor agent communication is the binding constraint.

• Talent. The scarcest resource in agentic AI deployment is people who combine AI engineering with life sciences domain knowledge and quality systems experience. Pistoia Alliance polls rank skills shortage as the second-biggest barrier to AI adoption in pharma, behind resistance to change. Many are trying to build these hybrid teams from scratch while simultaneously running pilots.

• Consolidation & Stratification. The field is splitting between companies with proprietary biological data and those building on public data alone. This matters because data moats increasingly determine which AI/agent systems can produce differentiated outputs.

The field is moving fast enough that a survey like this one dates quickly. There is a lot of inflated optics surrounding AI, and the current cycle is agents—so in the crossfire of major forces and infrastructure investments, that’s worth keeping in mind when gauging the reality.

The underlying tension between what these systems can do and what biology actually requires will stay for a while. We’ll be watching for named partnerships and plans, of course, but more so (and mainly) for tangible outputs and reproducible results.

As always, if you're working on any of this or watching it from the inside—we'd love to hear what you're seeing, leave a comment!

Leave a comment

⚠️ 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:

• Build virtual patients — AI integrates biomarkers, imaging, genetics, and real-world evidence to generate synthetic profiles capturing the full variability of real populations.

• Run simulated trials — virtual cohorts replace placebo groups or test experimental therapies in silico, probing efficacy and safety without exposing patients to unnecessary risk.

• Optimize continuously — trial parameters like dosing and sample size are continuously refined in real time, anchored by validation against real-world data.

A less computationally demanding synthetic control arm approach uses AI-generated patient data based on registries, and real-world evidence but unlike DTs not modelling it on a particular individual. The appeal is sharpest in rare diseases, where finding enough eligible control patients is often impractical. The FDA, EMA, and NICE have all endorsed the approach, and it’s gaining traction: recent Phase II/III myeloma and lymphoma trials have leaned on external control data, and in at least one case (blinatumomab for acute lymphoblastic leukemia), a synthetic control arm helped support accelerated regulatory approval. AstraZeneca used over 300M synthetic patient records to advance its clinical trials, allegedly saving up to $100M per drug in development.

Synthetic control arms built from historical data have already supported label expansions and accelerated approvals with alectinib, blinatumomab, palbociclib among them. AI-generated individualized digital twins, however, have not yet served as primary evidence in a completed approval.

🦾 An Industry Arm

Business models in this space vary significantly. Some companies license their platforms as SaaS tools to sponsors (Unlearn, Phesi), others embed digital twin capabilities within broader trial-management suites sold enterprise-wide (Medidata), and a few operate closer to a service model, generating synthetic data or external control arms on a per-study basis (ConcertAI).

The clearest commercial traction spans two approaches: prognostic covariate adjustment, which uses digital twins to shrink control arms within standard RCTs, and synthetic control arms, which replace part or all of the control group with external or simulated data.

⭐ Unlearn.ai’s PROCOVA method (EMA-qualified, with positive FDA feedback) embeds prognostic scores from digital twins directly into Phase 2/3 trial design. The company has worked with AbbVie and J&J on Alzheimer’s trials and partnered with Quralis and ProJenX for ALS, and raised $50M in a 2024 Series C. Notably, the EMA qualification covers the statistical adjustment procedure and applies only to trials with continuous outcomes, not the AI model-building step itself.

⭐ Phesi approaches the same problem from a data-scale angle, drawing on a clinical database surpassing 132M patients to construct digital twins that, as a milestone proof-of-concept study showed, can replicate standard-of-care arms in cGvHD trials—work that earned it recognition as a Frost & Sullivan Global Growth Leader in AI-enabled clinical trials.

⭐ Medidata, backed by Dassault Systèmes, offers synthetic control arm technology as part of a full-stack platform spanning 38,000+ studies; it recently extended its partnership with Sanofi for decentralized trials, and Anthropic’s Claude for Healthcare now includes a Medidata connector.

⭐ ConcertAI rounds out the trial-support layer, structuring real-world data from 11M+ patients into external comparator arms, now running on NVIDIA infrastructure after a 2024 integration agreement, and partnering with Bayer on multiomic cancer data.

💊 Drugs Before Trials

Before a trial even begins, a patient’s virtual replica can simulate treatment effects entirely in silico. That potential extends across every major stage of drug development:

• Early discovery: model disease mechanisms and surface therapeutic targets through biological simulation rather than empirical screening alone.

• Preclinical testing: simulate human responses to reduce reliance on animal models while generating more clinically predictive data.

• Clinical trial simulation: stress-test doses, treatment plans, and patient selection criteria on virtual cohorts before recruiting real ones.

• Regulatory submissions: supply in silico safety and efficacy evidence alongside clinical data, with review bodies increasingly weighing DT-specific concerns like model bias and algorithmic transparency.

• Post-market surveillance: continuously update with real-world data to monitor drug safety and efficacy after approval.

⭐ Aitia uses its causal REFS engine to model disease biology from multiomics data and simulate clinical outcomes entirely in silico, bypassing animal and cell-line models. The platform spans neurodegeneration, oncology, cardiometabolic disease, and immunology, validated by a 2024 partnership with Orion in oncology drug discovery, an extended agreement with Servier on brain cancer, and a 2025 partnership with Gustave Roussy to map biological causes of human cancer.

⭐ VeriSIM Life takes a pharmacokinetic angle: its BIOiSIM platform uses hybrid AI and mechanistic modeling to predict drug behavior in humans before trials begin, scoring compounds on a Translational Index (essentially a credit score for drug viability) across a search space of over one trillion compounds and 5,000+ validation datasets, claiming a 2.5-year average reduction in time to IND.

⭐ Also operating at the patient level: Orakl Oncology, a 2023 Gustave Roussy spin-off we covered previously, builds tumor avatars from patient samples to simulate drug responses and identify therapeutic targets, and raised €11M in 2024 seed funding.

🚧 Unavoidable Limitations

Despite obvious promise in clinical trials and drug development, digital twins face challenges as a technology. The Unlearn.AI’s team highlights four key limitations of the approach:

• Complexity and variability: Unlike mechanical systems, each patient has a unique biological profile shaped by genetics, environment, and lifestyle. Human physiology operates through dynamic, nonlinear interactions that are difficult to simulate — requiring novel AI architectures, robust validation procedures, and continuous model updating as standards of care evolve.

• Data availability and standardization: DT models must integrate health data from disparate sources. EHRs are abundant but inconsistent and heterogeneous, while clinical trial data is high-quality but limited in scale and population representativeness. Building globally representative datasets remains a massive challenge, compounded by data drift, changing measurement standards and socioeconomic factors. The input requirements for a credible trial-level twin (longitudinal biomarkers, imaging, genomics, lifestyle data) often exceed what trial sites currently capture at the point of care. Compounding this, clinical data standards remain fragmented: CDISC (FDA’s mandated submission format) was designed for traditional RCTs over twenty years ago and does not natively accommodate real-world data or non-interventional designs, while FHIR and OMOP offer better support for EHR and observational data but lack regulatory adoption as submission standards.

• Prognostic value and validation: Predicting individual responses to treatment is difficult given the complexity of biological interactions and frequent data gaps. AI models must handle missing data, diverse data types, and outcome distributions — all while maintaining clinical-grade precision. Robust, context-aware validation is essential but resource-intensive.

• Ethical and privacy concerns: Patient data is sensitive and heavily regulated. Key concerns span data collection, management, analysis, and use — including cybersecurity risks, algorithmic bias, and the need for transparent, continuously monitored models to ensure equitable outcomes.

🔭 Regulations and Lookahead

The regulations agencies are picking up the pace with respect to digital twins. In the US, while DTs are still not recognized as a tool, a 2024 collaboration between the NSF, NIH, and FDA is exploring this technology as a catalyst of biomedical innovation and their potential to transform preclinical and clinical research.

A Center for Drug Evaluation and Research (CDER) discussion paper, originally published in 2023 and updated in 2025, further outlined current and future applications of digital twins, highlighting their potential to accelerate drug development and support placebo arm replacement.

In case of a placebo control arm, no fundamental barriers exist: sponsors wishing to use a digital twin in place of a placebo control arm must notify the FDA at the IND filing stage and provide consistent, transparent updates throughout the trial.

In Europe, the EMA has taken concrete steps as well. In late 2022, its CHMP issued a Qualification Opinion for Unlearn’s PROCOVA. This was followed by the EMA’s five-year AI Action Plan, committing to technical deep dives into digital twin technology among other tools. Broader AI governance in the medical space is also addressed under the EU AI Act.

Checking back with our 2025 review, the European Commission’s Virtual Human Twins Initiative, backed by over €100 million in combined Horizon Europe/Digital Europe funding, has moved past the manifesto stage since last year. The EDITH coordination action published its strategic roadmap and policy brief in October 2025, and in June 2025 the Commission launched procurement for the VHT digital platform that will host model integration and validation. The Manifesto itself has grown to over 100 signatories by now, and the infrastructure is now being built.

Outside the US and EU, regulatory engagement is more tentative, a few instances:

• Japan’s PMDA has published an AI Action Plan and launched ‘Early Consideration’ publications to address emerging technologies in drug development, though it has not yet issued digital twin-specific guidance.

• In China, the NMPA’s 2022 Technical Review Guidelines for AI Medical Devices require that AI tools demonstrate data sufficiency, diversity, and representativeness, and mandate that self-learning algorithms be “locked” post-market unless resubmitted for review, a rules-based approach that some contrast with the more standards-oriented frameworks in the US and EU. Like Japan, China has not yet extended this framework to digital twin methodology as a distinct category.

Both agencies are modernizing, but neither appear to have matched the EMA’s qualification-level endorsements or the FDA’s explicit discussion-paper engagement with DT methodology.

Early examples like Phesi’s cGvHD work suggest that digital twins can already strengthen evidence generation in specific settings. The more plausible near-term future, however, is not a fully virtual clinical trial, but a hybrid model in which computational controls gradually replace or reduce traditional placebo arms where they are hardest to justify.

It should be noted we haven’t gone into questions like the commercial model, whether SaaS licensing, per-study pricing, or platform embedding will prove most durable, nor into how HTA bodies like NICE or IQWiG would weigh DT-generated evidence in reimbursement decisions. Both will shape adoption as much as the technology itself.

Shortly prior to last Christmas the European Commission published a proposal of a European Biotech Act, a strategic initiative aimed at setting up a regulatory framework to strengthen the life sciences sector across the EU. The document has been mostly positively received by the sector leaders as a needed step towards fostering local biotech innovation. Brussels isn’t stopping there. Commission President Ursula von der Leyen pitched EU-Inc, a pan-European company structure meant to solve what many see as the EU’s core startup problem of navigating 27 different bureaucratic regimes. The proposal would let one register in 48 hours, fully online and in English.

The Case for Urgency

Why does it matter? Europe gave the world its first blockbuster pharmaceutical (Aspirin in 1899) and just 30 years ago produced half of all new treatments globally. Today, that share has fallen to just one in five. Even though the EU biotech industry has grown twice as fast as the overall union’s economy over the last decade, it struggles to convert the world’s top science into commercially viable products.

Europe holds a comparable share of the top 10% most-cited biomedical research to the US and China, yet lags significantly behind in venture investment — a gap caused by underdeveloped private equity markets and fragmented, complex regulatory frameworks. The disparity is also visible in listing trends, with 66 of the 67 EU companies that went public over the past six years choosing foreign stock exchanges.

To address this, the Biotech Act includes measures like:

Read more

Transformer architectures trained on biological data are beginning to predict drug response, generate therapeutic hypotheses, and identify which patients are likely to benefit from which treatments (part of a broader push that includes early attempts at virtual cell models) tasks that previously required years of wet-lab iteration. Some of that work is still early, though a handful of results have made it far enough through validation to be worth paying attention to.

• Google Research, Google DeepMind, and Yale spent much of 2025 scaling C2S-Scale, a language model that reads single-cell RNA data as text; the 27-billion-parameter version, released in April, came in October with wet-lab validation of a model-generated hypothesis about making immune-”cold” tumors visible to T cells.

• A collaboration between Microsoft Research, Providence Health, and the University of Washington took a complementary approach: GigaTIME, published in Cell in December, routinely converts pathology slides into virtual immune-protein maps, surfacing over 1,200 significant associations across 14,256 patients.

• At Davos in January, Demis Hassabis now put Isomorphic Labs‘ first trials, primarily oncology candidates, at end of 2026; the company followed this month with IsoDDE, a general-purpose drug design engine that reportedly doubles AlphaFold 3’s accuracy, already deployed across its oncology programs.

Not all of it is language-model work.

Read more

Just a day earlier, MIT Technology Review published its annual 10 Breakthrough Technologies list. This year, three of the highlighted technologies were in genomics: personalized gene editing, embryo scoring, and gene resurrection. From there, it seems like genomic applications are moving more into the mainstream technology discourse.

Another just-in data point from a few days ago is a report out of San Diego, where Element Biosciences says its newly announced VITARI benchtop sequencer can deliver a whole genome for $100, positioning it as a lower-cost alternative to Illumina’s high-throughput systems.

Looking at these and many of last year’s developments, genomics come into view as an integrated technology wave that extends from data generation to interpretation, intervention, and biological reconstruction.

With those early-2026 pings as a starting point, let’s do a selective pass through a few genomics patterns that seem to be carrying momentum into 2026.

Read more

At first glance, these headlines span different companies and medical areas. But they share a common thread: each centers on advanced therapeutic modalities (ATMs)—a new generation of medicines that go beyond the limits of conventional drugs.

According to BCG, eight of the top ten best-selling biopharma products in 2025 are new-modality drugs, and the global pipeline value for these therapies has reached $197B. ATMs are becoming a more established part of the industry and are noticeably contributing to its growth.

Reject Tradition, Embrace Modernity

Read more

Read more

In DealForma’s figures cited by CEO Chris Dokomajilar, deal flow between large-cap biopharma and Chinese biopharma accelerated in 2024-2025. In 2025, big pharma completed 18 in-licensing and asset purchase deals (just one in 2020) from Chinese companies with $50M+ upfronts, totaling $57.3B in deal value and $3.9B in upfront cash and equity. By 2026, China continues to emerge as a major source of globally licensable, clinical-stage biotech assets, backed by an increasingly complete innovation stack, even as new policy constraints complicate cross-border data flows and outsourcing.

In late January, speaking at the Asian Financial Forum in Hong Kong, executives from Merck and Amgen pointed to China as a likely early approval market for fully AI-designed drugs. Merck China president Marc Horn suggested that 2026 could mark the shift from AI-assisted discovery to compounds designed end-to-end by AI entering regulatory pipelines, citing China’s patient datasets, clinical execution, and the government’s recent “AI Plus” policy push. Amgen’s chief medical officer Paul Burton pointed to a similar timeline, seeing 2026 as a year when AI-driven and human genetics–led discovery could begin translating more directly into drug candidates.

For perspective, among recent big pharma deals involving Chinese companies, this year’s JPM week had AbbVie’s $5.6B partnership with RemeGen around a bispecific oncology asset. Looking back at just 2025, Pfizer licensed a bispecific from 3SBio with $1.25B upfront, AstraZeneca entered a multi-year $5.3B AI-enabled small-molecule discovery collaboration with CSPC Pharmaceuticals, and GSK’s x Jiangsu Hengrui agreements included $500M upfront and up to about $12B in potential milestones.

From Generics to Innovation

Read more

Read more

Invasive & Minimally Invasive BCIs

Brain-computer interface (BCI) systems are being explored and used as a way to restore lost motor, speech, or sensory functions, particularly in patients with paralysis or neurodegenerative conditions. They work by placing electrodes on or in the brain to capture high-resolution neural activity, which is then translated into actions like moving a cursor, generating speech, or triggering stimulation.

Typically, BCIs include implanted pulse generators and wireless connections to external processors, which decode brain signals such as spikes or local field potentials from targeted brain areas, then use trained algorithms to translate those activity patterns into outputs such as cursor motion, text, or stimulation commands.

In 2025, several programs moved into multi-center or early pivotal territory:

• Neuralink extended its PRIME program into Great Britain with the GB-PRIME study at UCLH and Newcastle, evaluating the fully implantable N1 interface in patients with motor neuron disease and spinal cord injury, and reporting the first UK patient controlling a computer within hours after surgery. The same implant was used at home by ALS patient Brad Smith to control a motorized Insta360 webcam, demonstrating extended real-world use beyond cursor control.

• Paradromics received FDA IDE approval for its Connexus system to start the Connect-One early feasibility study, targeting speech restoration and computer control in people with severe paralysis via a high-bandwidth, fully implantable BCI. The Connect-One trial is designed around speech restoration as a primary endpoint rather than generic cursor control.

• Precision Neuroscience advanced its thin-film Layer 7 cortical interface. The 1,024-electrode subdural array, FDA-cleared as a temporary mapping device, was profiled in first human recipients as a minimally invasive, high-density platform that sits on the cortical surface rather than penetrating tissue.

• CorTec’s Brain Interchange BCI system reached first-in-human use in a stroke patient as a fully wireless, closed-loop implant capable of recording and stimulating cortex in real time, positioning it as a European competitor in implantable neuromodulatory BCIs.

• Synchron introduced an updated version of its endovascular Stentrode BCI that integrates Nvidia AI and the Apple Vision Pro headset to let people with severe paralysis control digital and physical environments using neural signals. Later, Synchron publicly demonstrated a person with ALS using its implanted Stentrode to control an iPad entirely by thought by converting neural motor-intent signals into native iPadOS inputs.

Read more

There are other speakers highlighting the promises of AI for solving aging. Anthropic CEO Dario Amodei said at 2025 WEF that if AI dramatically accelerates biological research, doubling the human lifespan by around 2030 isn’t unrealistic because it could compress “100 years of progress” into 5–10 years. Such claims are controversial, but they reflect a real trend: AI is impacting both basic geroscience and emerging longevity medicine. Before delving deeper into the intersection of AI and longevity, let’s overview the history of this field before machines came.

Nothing Lasts Forever

Aging is the gradual, time-dependent decline in the physiological functions required for survival and reproduction. Unlike age-related diseases (such as cancer or heart disease), the defining features of aging are shared by all individuals within a species.

As an integral part of life, aging has caused a multitude of philosophical disputes throughout history, tracing back to 350 BCE when Aristotle first tried to explain senescence, viewing it as a ‘natural illness’. However, conventional aging research started much later, in the 20th century.

Read more

Terray’s work in diffusion methods prompted a broader reflection on generative AI in biology. Today, most conversations and publications center on Transformer-based systems, especially large language models (LLMs) and other foundation models (FMs). LLMs make up a major subset of FMs, but whereas language models are trained primarily on textual data like natural language, code, or biological sequences, foundation models extend the paradigm to additional modalities, including images, audio, video, and even multimodal combinations.

Recent meta-reviews in biomedical NLP collectively catalog nearly 300 LLM instances across hundreds of studies. Foundation models are also proliferating, with over 200 tools developed since 2022 in drug discovery alone. In contrast, the literature on diffusion models for biological and chemical applications remains comparatively modest. So far, there have been only a handful of reviews capturing the diffusion generators. Yet despite lower popularity, diffusion architectures are carving out a meaningful and distinctive role in biotech research and industry.

Before diving deeper into their role in biomedicine, let’s briefly review how diffusion models work in general.

Read more

This transformation is happening thanks t…

Read more

EvolutionaryScale emerged in 2023 after Rives, along with Tom Sercu and Sal Candido, left Meta’s AI protein group (FAIR) during the company’s “year of efficiency” (there are, again, plans to cut 600 AI jobs after a $14.3 billion Scale AI investment and hiring spree this summer). Backed by the likes of Amazon and Nvidia, the team raised $142 million to develop large-scale generative models for protein design and became known for the ESM family of protein language models (PLMs) trained directly on amino-acid sequences.

Its flagships, ESM3 and ESM Cambrian, extended this work to fully generative modeling of protein structure and function. ESM3, trained on 2.7 billion proteins, has already been used to design molecules like the novel green fluorescent protein variant, esmGFP, said to represent roughly 500 million years of natural evolution.

CZI’s Biohub folds this hire into its broader “virtual biology” plan, setting out four scientific challenges: building an AI-based model of the cell, advancing imaging, instrumenting inflammation, and using AI to reprogram the immune system, with the Virtual Immune System as one of the flagship projects. The ES team is brought in “to help advance this initiative.” In the VIS roadmap, the molecular-interactions axis explicitly calls for protein language models that “can learn the universal grammar of immune recognition and enable the rational design of novel receptors.”

With that, let’s step back and look closer at what protein language models are, what kinds of applications companies are building them for, and where pharma is already involved.

Read more

Last year, OpenAI partnered with Formation Bio and Sanofi as well as signed agreements with Moderna, Eli Lilly; followed by a Thermo Fisher Scientific deal in 2025. At the same time xAI advertises Grok for Government with support for science and healthcare purposes, while DeepSeek gains adoption across Chinese hospitals.

Today we’ll look at LLMs entering biomedical workflows, examine what these systems can do in lab- and clinic-adjacent tasks, and how hybrid designs aim to mitigate common failure modes.

Read more