Top market ‘AI in biotech’ stories
Here’s your latest installment of our selection of AI in biotech news.
In just 3–4 minutes, catch up on the innovations reshaping drug discovery and biomedical research. We are delivering fresh insights to your inbox on the second Tuesday of each month, so you can stay informed and inspired. Happy reading!
In today’s edition:
🧬 ML platform for gene-disease mapping licensed to AstraZeneca
🖥️ Eli Lilly builds AI supercomputing for end-to-end drug development
🧠 Google AI generates a cancer hypothesis – and it holds up in wet-lab validation
💡 Generative model designs antibodies de novo with cryo-EM validation
AstraZeneca Signs a $555M Deal with AlgenBio
AstraZeneca has licensed AlgenBrain, a machine-learning platform developed at UC Berkeley’s Jennifer Doudna Lab. The system maps gene function to disease outcomes, offering a computational route to identifying novel targets in immune-related disorders.
The agreement is valued at up to $555M, reflecting the strategic importance of AI-derived insights in target selection and early translational research.
AlgenBrain integrates large-scale perturbation data with ML models to uncover mechanistic relationships that are otherwise difficult to detect.
source: Reuters
Eli Lilly Builds an AI Supercomputing Infrastructure for End-to-End Drug Development
Eli Lilly is partnering with NVIDIA to deploy a DGX SuperPOD, built specifically to scale AI across discovery, development, manufacturing, and imaging workflows.
The system will allow Lilly to train models on millions of experimental outcomes, expanding the therapeutic search space far beyond conventional screening.
A subset of these models will be shared through Lilly TuneLab, a federated platform enabling biotechs to use Lilly-trained models without exposing proprietary data.
source: Reuters
AI Generates a Cancer Hypothesis – and It Gets Validated
Google’s foundation model C2S-Scale 27B, trained on more than a billion single-cell profiles, has generated a previously unknown hypothesis about tumor immune evasion.
The model predicted that silmitasertib could enhance immune recognition of “cold tumors,” a class of cancers typically resistant to immunotherapy.
Follow-up experiments in human cells validated the prediction – demonstrating that the model was not retrieving known patterns, but inferring mechanistic relationships from large-scale cellular data.
C2S-Scale interprets molecular signals as “cell sentences,” enabling it to reason across perturbation landscapes and propose biologically testable ideas.
source: ET Edge Insights
Designing Antibodies de Novo: RFantibody Opens a New Chapter for Biologics
A new Nature paper from the Institute for Protein Design introduces RFantibody, a generative AI model that designs antibody sequences de novo and predicts folding and binding prior to experimental testing.
Key results from the initial evaluation:
• Mid-nanomolar binding
• 4 out of 5 AI-designed antibodies matched predicted structures (validated with cryo-EM)
• Code released openly to support community-driven benchmarking and extension
The work demonstrates how antibody discovery may shift from empirical screening toward intentional, model-guided design, compressing early R&D timelines and increasing design precision.
source: Financial Times
