How 2026 Started: First-Weeks Readout on AI, Pharma, & Policy

Pharma interest in virtual-cell-style modeling showed up in two directions: GSK signed a five-year, non-exclusive license and collaboration with Noetik, committing $50M to use its virtual cell foundation models and bespoke spatial oncology datasets in non-small cell lung cancer and colorectal cancer, while Bristol Myers Squibb entered a strategic clinical-development agreement with Immunai to apply its immune “operating system” to high-dimensional trial data to decode patient immune responses/variability in treatment outcomes.

Another clear theme was the continued movement of AI from proofs-of-concept into everyday workflows. Pharmaceutical players started bringing AI in-house and embedding it into R&D operations – AstraZeneca’s mid-month acquisition of Modella AI was explicitly to fold foundation models and “agentic” AI workflows throughout its oncology pipeline. On the tools side, MIT spinout Boltz (the founder trio is behind the well-known BoltzGen model) launched as a public benefit corporation with $28M seed round, and a multi-year collaboration with Pfizer.

Below is a selective rundown of the early-year signals we tracked most closely.

💊 AI in Drug Discovery

Among some of the notable platform partnerships and deployments, Isomorphic Labs partnered with Johnson & Johnson to apply its platform across multi-modality drug design in hard-to-drug indications, while pushing back its own AI-designed drug trials to end-2026. Servier also partnered with Iktos to use its AI-and-robotics DMTA platform for small-molecule discovery.

As for software integration, Insilico Medicine (IPO’d in Hong Kong with >$290M just at the end of last year) integrated its drug discovery foundation model into Microsoft Discovery and launched “Science MMAI Gym” to train and evaluate LLMs on pharma R&D tasks.

🔹 AI Drugs in Clinic — AI-originated assets continued to progress with several notable selective clinical touchpoints. Insilico Medicine advanced a gut-targeted PHD inhibitor into Phase II for ulcerative colitis, secured an IND for a brain‑penetrant NLRP3 inhibitor in CNS disease, and nominated a preclinical small molecule with Hisun eight months into collaboration. Lantern Pharma received a third orphan designation for its AI-designed LP‑284.

🔹 Post-AlphaFold — ByteDance, the Chinese company behind TikTok, introduced a protein folding system building on AlphaFold 3, alongside the emergence of “post-folding” approaches like DrugCLIP, which embeds proteins and small molecules into a shared representational space.

🔹 Against developability constraints — New generative models started to encode real-world constraints into sequence design: 1910’s PEGASUS is designed to target cell-permeable macrocyclic peptides, addressing a long-standing challenge in peptide drug development—designing molecules that retain favorable drug-like properties while still being able to enter cells; and Absci’s launched Origin-1, focused on de novo antibody design against difficult targets.

DeepMind’s AlphaGenome, first previewed in 2025, and models regulatory effects of non‑coding variants across long DNA windows, framing the AI tool potentially suitable for rare-disease variant interpretation and synthetic DNA design.