Weekly Tech+Bio Highlights #74: Big Pharma AI Tie-Ups Lean Toward Real-World Data
Quick run of pharma-AI collaborations, a new startup based on Google's cell sentence tech, and $100 genome sequencing from San Diego
This past week’s pattern was a cluster of big-pharma and large healthcare collaborations that pull AI closer to multimodal biological and clinical data, and closer to lab and development workflows, with several deals pointing at “model plus measurement” loops.
Separately, a new benchtop sequencer announcement from San Diego kept the “falling sequencing costs” thread in view, which we framed in our recent ‘genomic watchpoints for 2026’.
We also saw a fresh Y Combinator entrant building an agentic drug platform on Google DeepMind’s Cell2Sentence single-cell foundation model, and compiled a short “New AI-Native Biotechs” list for a quick scan of new launches.
Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.
If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:
This newsletter reaches over 11K industry professionals from leading organizations across the globe. To reserve your sponsor slot in one of the upcoming issues, contact us at info@biopharmatrend.com
🤖 AI x Bio
(AI applications in drug discovery, biotech, and healthcare)
🔹 Insilico Medicine shared its new ACS Central Science “Prompt-to-Drug” paper, outlining a “pharmaceutical superintelligence” architecture that orchestrates agentic AI plus automated labs to turn simple prompts into drug candidates—positioned as a foundation for its $100M partnership with Lilly.
🔹 Deep Origin and Arctoris were selected for the UK ARIA AI Scientist program to build an AI-enabled autonomous scientist for endometriosis, combining Deep Origin’s computational drug discovery with Arctoris’ robotic lab to iteratively plan, run, and analyze experiments with minimal manual input.
🔹 Flatiron launched AI-powered prostate cancer datasets spanning ~400,000 patients across the US, UK, and Germany.
🔹 Integrated Biosciences added Tony Wu (xAI co-founder) to its Scientific Advisory Board to strengthen its optogenetics-chemistry-AI platform for discovering small-molecule therapies for age-related diseases.
🔹 UK-based Bioptic published a paper on an agentic AI system for drug asset scouting that analyzes scientific, regulatory, and corporate sources to flag partnership targets and competitive risks, and reports early pharma BD adoption with a public specialty therapeutics company using it for diligence and portfolio intelligence.
🔹 Tran et al. (Science, Feb 2026) introduced MULTI-evolve, a lab-in-the-loop ML method that uses protein language models plus targeted double-mutant testing to learn mutation interactions and quickly design stronger multi-mutation proteins, delivering big gains across APEX enzymes, dCasRx, and an anti-CD122 antibody in fewer iteration cycles than traditional directed evolution.
💰 Money Flows
(Funding rounds, IPOs, and M&A for startups and smaller companies)
🔹 PitchBook expects biopharma VC to keep recovering in 2026 (report PDF), but in a more cautious, selective way: money is concentrating into fewer, bigger rounds for more proven, later-stage drug assets rather than early platform bets (2025 deal value: $33.8B; median early-stage round: $27M, up 77%).
🔹 After announcing upcoming Phase 3 data from its psilocybin trials in treatment-resistant depression, Compass Pathways launched a proposed $150M public offering, with an option for underwriters to purchase up to an additional $22.5M.
🔹 Korsana Biosciences emerged from stealth with $175M raised to advance its brain-shuttle platform, including lead Alzheimer’s candidate KRSA-028, a shuttled anti-amyloid beta antibody partnered with Paragon Therapeutics, with plaque-clearance data expected in late 2027.
🔹 Innovacell, a Tokyo-based cell therapy company, plans to IPO on the Tokyo Stock Exchange on Feb. 24, 2026, aiming to raise about ¥14.17B (~$92M) to support its Phase 3 program and build manufacturing/commercial infrastructure.
🚜 Market Movers
(News from established pharma and tech giants)
🔹 Unnatural Products (UNP) signed a research collaboration and licensing deal with Novartis to develop AI-designed macrocyclic peptide therapeutics for historically “undruggable” targets (with potential cardiovascular applications), with UNP eligible for up to $100M upfront/pre-IND, up to $1.7B in milestones, plus tiered royalties.
🔹 BostonGene and Daiichi Sankyo struck a strategic collaboration to use AI-driven multimodal “digital twin” analytics (multi-omics + histopathology) to support precision ADC development.
🔹 Merck and Mayo Clinic formed an R&D collaboration to combine Mayo’s de-identified multimodal clinical/genomic data and platform tools with Merck’s AI/ML capabilities to accelerate AI-enabled drug discovery and precision medicine, initially focusing on IBD, atopic dermatitis, and multiple sclerosis.
🔹 GE HealthCare expanded its BARDA partnership by ~$35M to advance AI-powered ultrasound tools for faster trauma assessment and mass-casualty emergency care.
🔹 Scribe Therapeutics and Eli Lilly hit a second milestone in their multi-program in vivo CRISPR collaboration for neurological and neuromuscular disorders, powered by Scribe’s X-Editor (engineered CasX) platform; financial terms weren’t disclosed, but the 2023 deal includes $1.5B+ in potential milestones plus royalties.
🔹 Boehringer Ingelheim has stopped testing and dropped its inhaled gene therapy for cystic fibrosis after an early clinical trial didn’t show results strong enough to keep going, even though it appeared reasonably safe.
⚙️ Other Tech
(Innovations across quantum computing, BCIs, gene editing, and more)
🔹 Ultima Genomics launched its UG200 Series DNA sequencers and Solaris 2.0 workflows ahead of AGBT today, claiming up to 2x more output and ~50% faster runs than UG100, with shipments starting Q2 2026 and a Solaris 2.0 Max chemistry later in 2026 targeting up to 20B reads per wafer.
🔹 San Diego-based Element Biosciences unveiled a benchtop DNA sequencer that can deliver a whole genome for $100 (allegedly ~half the cost of Illumina’s NovaSeq) intensifying competition as the two companies battle in court over antitrust and patent claims.
🔹 An FNIH Biomarkers Consortium study in Nature Medicine found a “clock model” using a single blood p-tau217 test can predict Alzheimer’s symptom onset about 3-4 years in advance, alongside a new web tool to visualize biomarker trajectories for improved trial planning.
🏛️ Bioeconomy & Society
(News on centers, regulatory updates, and broader biotech ecosystem developments)
🔹 The FDA is shifting its default drug-approval standard to one well-controlled pivotal trial, with additional “confirmative evidence” (e.g., validated biomarkers, mechanistic data, and real-world evidence) expected to support the application.
🔹 The Pentagon briefly added major Chinese firms (including WuXi AppTec) to its Section 1260H “Chinese military companies” list ahead of an expected Trump-Xi summit, then withdrew the Federal Register filing without explanation, leaving the designations in limbo and adding fresh uncertainty to US-China tech tensions.
🔹 A study tested whether chatbots help beginners do key lab steps needed to make a virus; surprisingly, people with LLM access did not do better than those using the internet alone (about 5% vs ~7% completed all tasks). Forecasters (especially “superforecasters”) had expected a much bigger boost, and after seeing the results they slightly lowered their estimates of near-term biosecurity risk, while noting that a future big jump in AI lab help could still raise risk.
🚀 A New Kid on the Block
(Emerging startups with a focus on technology)
🔹 CellType, a Yale University spinout, joined Y Combinator’s W26 (2026) batch to develop an agentic AI drug platform based on Google DeepMind's Cell2Sentence, a 27B-parameter single-cell foundation model that treats gene expression as a biological language to simulate human biology in silico.
Read also:
Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape
Great roundup. The 'model plus measurement' framing is the right lens here. Most commentary focuses on the AI angle, but the real play is acqui-hiring access to rare longitudinal multimodal datasets that took decades to assemble. Having followed these deals closely, I think the FDA's shift toward single-pivotal trials with confirmtory real-world evidence is what makes these parterships strategically different: it's turning data moats into regulatory moats.
The $100 whole genome keeps coming up in lectures as the threshold where sequencing stops being a specialist tool and becomes a routine clinical one. But I wonder whether the bottleneck just shifts at that point. Most hospitals do not have the bioinformatics infrastructure to act on a whole genome meaningfully. Does cheaper sequencing accelerate that investment, or does it just create a larger backlog of data nobody knows what to do with?