In Focus on Life Sciences Digital
On 27 October, Anthropic reported early results from research partnerships with six drug developers: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. At Generate, six scientists used Claude Code to propose redesigned components for CAR T-cell therapy aimed at solid tumours. BMS is using Claude to review the clinical trial plans for one of its early-stage drugs and estimate how likely those trials are to succeed. The two companies plan to expand this work in 2027. The Column Group is forming a new company, Pleiades, to run its drug programmes with Claude.
In the same update, Anthropic disclosed that it had acquired Mithril Medicine, a one-person biotech whose founder, Cody Tranbarger, used Claude to coordinate contracts, logistics and experiments across 11 external labs. Tranbarger took a candidate for a rare muscular dystrophy to promising mouse results in under 18 months, and human trials could start as early as late 2028. A day earlier, Anthropic’s S-1 named its molecular biology lab and the Coefficient Bio acquisition among its life sciences investments and set out the regulatory and operational risks that come with them.
OpenAI moved on the operational side. On 20 October it acquired Orchestra Bio, a platform where scientists and programme leaders delegate study monitoring and tracking to AI agents. OpenAI is folding it into Rosalind, its life sciences suite already used by Moderna and Novo Nordisk. When a manufacturing batch fails, Orchestra’s agents trace which studies and timelines are affected and lay out options for the people who decide. Terms were not disclosed.
Both companies now want more than to sell models to pharma. They want to run parts of drug development themselves, from planning and coordinating programmes to, in Anthropic’s case, developing its own drugs. For pharma and biotech teams, picking an AI vendor is becoming closer to picking a development partner, with real questions about who owns the data and who carries the risk when an AI agent suggests the next step. Companies selling R&D planning and trial design software now compete directly with OpenAI and Anthropic.
Also Recently:
On 23 October, Pathos AI closed a $237 million Series E at a post-money valuation of about $3 billion, led by Bristol Myers Squibb. In return, BMS gets access to Foundry, Pathos’s AI platform for drug development, across its R&D portfolio, plus three years of exclusive use in neuroscience. Tempus will build a de-identified dataset from hundreds of thousands of patients with neurological conditions and license it to Pathos for $55 million over five years. Capital for AI-native drug developers kept coming: Formation Bio launched a $600 million fund to acquire 10 to 15 clinical-stage programmes over three to five years, and Flagship Pioneering merged Cellarity, Etiome and Empress Therapeutics into Etcelera.
Dassault Systèmes has completed its acquisition of ArisGlobal for $1.8 billion in cash, with up to $200 million in additional consideration. It is now bringing its life sciences and healthcare capabilities together in 3D UNIV+RSES, one AI-native platform spanning discovery, clinical development, manufacturing and patient outcomes. Ashwini Zenooz, previously at the US Department of Veterans Affairs and Salesforce, takes the newly created role of CEO of the Life Sciences and Healthcare sector. With Medidata, BIOVIA and now ArisGlobal’s safety and regulatory software, Dassault holds one of the few portfolios that covers the whole drug lifecycle. The test is whether customers adopt it as one connected system.
On 29 October, Parse Biosciences, a QIAGEN company, and Nomic Bio announced Cascade. It is a licensable perturbation dataset that will pair 100 million single-cell RNA-seq profiles with more than 11,000 proteomic secretome profiles across over 150,000 cytokine and drug perturbation conditions. GSK, Pfizer, Sanofi, Graph Therapeutics and Turbine have signed on as foundational partners. Three weeks after Biohub’s $1.8 billion open data commitment, Cascade shows the commercial side of the same bet: pharma is willing to pay for proprietary, AI-ready perturbation data to train virtual cell models.
Tool Spotlight from our Life Sciences Digital database
Biomni Lab by Phylo
RESEARCH INTELLIGENCE AND DISCOVERY
Biomni Lab is Phylo’s agentic platform for biomedical research. It connects more than 300 databases, software tools and analytical methods, so scientists can search literature, analyse omics data and plan experiments in one workspace. Scientists can check the code and sources behind every result.
On 22 October, Bayer signed up to use Biomni Lab in drug discovery, so its scientists can pull together published research and omics data in one place.
🚀 Your solution, in front of the people building the future of life sciences.
Signals & Market Moves
The Lab Loop Gets Its Infrastructure 🔗
On 21 October at GTC Berlin, NVIDIA and the UW Medicine Institute for Protein Design released RFD4-Proteina, an open model that designs a protein’s sequence and all-atom structure together. In one reported lab test, 158 of 188 designs bound the target. Genedata, a Danaher company, added the model to its production discovery workflows the next day, and on 28 October it launched an electronic lab notebook built into its platform, so experiment records stay linked to molecules, samples and assay results. Opentrons raised a $40 million Series E to expand its lab robots and software, which AI systems can operate directly. More than 11,000 of its robots are already in labs in over 80 countries.The signal: The design model is becoming the cheapest part of the stack. Open weights for protein design are available through about ten platforms, so value moves to whoever closes the loop between design, robotic execution and a clean scientific record. Danaher, Genedata and Opentrons are each building that layer, which follows directly from last issue’s autonomous lab story.
On 22 October, Boehringer Ingelheim and Microsoft agreed to research digital patient twins for disease-progression modelling, trial simulation and synthetic or hybrid control arms, starting in lung health. On 20 October, AstraZeneca signed a non-binding agreement with Oracle to use real-world data and AI to identify eligible trial patients earlier, beginning with a hypertension programme with M42 in Abu Dhabi, and Syneos Health partnered with Optum on adaptive development planning that simulates trial scenarios before and during a study.
The signal: The scarcest input in trial design is data about real patients, and it sits with EHR vendors, payers and health systems. These deals pair sponsors and CROs with partners who hold it. Most of them are still exploratory, and Boehringer said any use of twins would need rigorous validation and regulatory acceptance. The milestones to watch are the first protocol built on a twin-informed control arm and recruitment numbers from the Oracle pilots.
On 23 October, NICE published draft guidance recommending five AI tools for detecting prostate cancer on MRI for NHS use during a two-year evidence-generation period, and declined two others for lack of evidence. Australia’s TGA found that 8 of 12 digital scribes it reviewed had functions likely to meet the medical device definition, although no supplier classed its product as a device. The FDA finalised its guidance on prescription drug use-related software, treating the output of sponsor-distributed companion apps as drug labelling.
The signal: In all three cases, what the software actually does counts for more than how the vendor describes it. The Australian scribe suppliers did not consider their products medical devices, and the TGA found that most of them likely are. For AI vendors, evidence plans and a clear statement of intended use now belong in the product roadmap from the start. NICE’s approach, conditional use paired with structured data collection, is likely to spread to other tool categories.
Events & Calls
Anthropic Guest Investigator Program, Life Sciences
For principal investigators at academic, nonprofit and government labs: 8 to 12 weeks working with Anthropic researchers at no cost, with investigators keeping the rights to their data. Applications opened on 22 October for about two weeks, so the window closes in the coming days.
Poreva Open: Global AI Protein Design Challenge
MegaRobo’s competition will put more than 10,000 AI-generated protein designs through standardised, double-blind wet-lab validation across five challenges. Registration is open, with first submissions planned for November.
BIO-Europe 2026, Cologne, 9 to 11 November
Europe's flagship biopharma partnering conference, with regulatory, clinical and manufacturing tracks. Digital partnering follows on 17 to 18 November.
HLTH USA 2026, Las Vegas, 15 to 18 November
The largest US health innovation event, and a good place to see how pharma companies and app developers respond to the FDA’s new guidance on drug-related software.
MEDICA 2026, Düsseldorf, 16 to 19 November
The world's largest medical technology trade fair and the central European meeting point for health IT, laboratory technology and diagnostics. We will be attending, so reach out to say hello! 👋
🤝 Got a tool for life sciences you'd like more people to know about?
Submit your tool here.
Want to be featured in a future issue or explore sponsorship?
We highlight the AI tools, companies, and initiatives shaping the future of life sciences. If you'd like to collaborate, reply to this email or get in touch: [email protected]
