DIGITAL BIOLOGY
Features
Takeda partners with Insilico to use Pharma.AI across its drug discovery pipeline
The collaboration spans molecule design through candidate selection across Takeda's therapeutic areas. Insilico will receive approximately $60 million in initiation fees, near-term payments and milestones. Success-based payments could bring the deal to approximately $600 million, plus royalties. Takeda receives exclusive worldwide rights to develop, manufacture and commercialize any resulting therapeutics.
Anthropic opens Claude Science in beta with more than 60 scientific skills
The workbench connects researchers to tools for genomics, structural biology and other scientific workflows. It runs on local computers or research computing systems, allowing datasets to remain local while sending the analysis context needed for each step to Claude. A reviewer agent checks citations, calculations and code. The beta is available to Pro, Max, Team and Enterprise users.
Xellar raises $50 million to connect automated experiments with biological AI models
The Series A and A+ financing will expand Xellar's organ-on-chip systems, laboratory automation, imaging and multi-omics capabilities. The company combines these tools to generate human-relevant experimental data for biological modeling. It plans to strengthen its computational biology teams and develop virtual cell technologies. Predicting therapeutic responses remains an ambition of the platform, not a clinical outcome established by this financing.
Mankind Pharma partners with Denovo Sciences on AI-led drug discovery
The programme combines Denovo's molecular generation and prioritization platform with Mankind's experimental validation infrastructure and clinical development expertise. Scientists will guide and validate the AI's proposals under a human-in-the-loop approach. The partners aim to shorten early discovery and improve lead selection; the announcement does not report measured gains or name specific drug targets.
MGI and Shanghai AI Laboratory introduce agents for automated laboratory execution
ProtoPilot connects protocol design, device-specific code and wet-lab feedback, while BioLab Bench evaluates whether agents can translate experimental requests into executable operations. MGI reports a 52.38% score on the ProtocolQA reasoning benchmark, close to its cited 54% human-expert reference. Those scores concern protocol reasoning, not autonomous laboratory success. The announcement builds on a June preprint.