Structure Prediction

6 articles


FIRSTAI-designed drug enters human trialsAI DESIGNPRECLINICALPHASE IIsomorphic LabsBuilt on AlphaFold + IsoDDE$3BLilly + Novartis partnerships
Clinical Trials

DeepMind CEO Confirms First AI-Designed Cancer Drug Entering Phase 1 Clinical Trials in 2026

Demis Hassabis confirmed that Isomorphic Labs' first AI-designed cancer drug will enter Phase 1 clinical trials in 2026, built on the AlphaFold-derived IsoDDE platform and backed by $3 billion in Lilly and Novartis partnerships.

·2 min
AlphaFold 3IsoDDEProtein-ligand prediction2× more accurateAntibody-antigen structures2.3× vs AF3, 19.8× vs BoltzNovel binding pocketsfrom sequence aloneBinding affinitybeats FEP at fraction of costIsomorphic Labs · February 2026
Structure Prediction

Isomorphic Labs Unveils IsoDDE Drug Design Engine, Doubling AlphaFold 3's Accuracy in Protein-Ligand Prediction

Isomorphic Labs unveiled IsoDDE, an AI drug design engine that more than doubles AlphaFold 3's accuracy in protein-ligand prediction and identifies novel binding pockets from sequence alone, as the company prepares for its first clinical trials.

·2 min
OPEN SOURCE1,000,000base pairs · single-nucleotide resolution3,000scientists using it160countries25/26benchmarks won
Foundation Models

Google DeepMind Open-Sources AlphaGenome, an AI Model That Predicts DNA Function Across 1 Million Base Pairs

Google DeepMind open-sourced AlphaGenome, an AI model published in Nature that predicts DNA function across 1 million base pairs, already used by nearly 3,000 scientists worldwide for cancer and disease research.

·2 min
Building open datasets forAI drug discovery€60MPFESGC18partners9countries5-year project · Led by Pfizer + SGCProtein-ligand binding data made freely availablefor the entire research community
Virtual Screening

LIGAND-AI Consortium Launches with 60 Million Euros to Build Open AI Drug Discovery Datasets

The LIGAND-AI consortium launched with over 60 million euros in EU funding, bringing 18 partners led by Pfizer and the SGC together to generate open datasets of protein-ligand interactions for training AI drug discovery models.

·2 min
Nature MethodsBaker Lab · U. of WashingtonDesigning functional enzymes from scratch41/41active sites scaffolded (vs 16 with previous methods)53kM⁻¹s⁻¹ catalytic efficiencyApproaches natural enzyme activity levelsRFdiffusion3 also released: 10× faster, open sourceJanuary 2026
Structure Prediction

RFdiffusion2 Designs Functional Enzymes from Scratch, Published in Nature Methods

David Baker's lab published RFdiffusion2 in Nature Methods, demonstrating an AI model that designs functional enzymes from scratch with catalytic activities approaching natural enzymes, alongside the open-source release of the faster RFdiffusion3.

·2 min
CHAI DISCOVERY · SERIES B$1.3Bvaluation · $130M raisedChai-2: de novo antibody design100×vs previous methods16%zero-shot hit rate86%drug-like profilesBacked by OpenAI, Thrive, General Catalyst
Antibody Design

Chai Discovery Raises $130M Series B at $1.3B Valuation, Unveils Chai-2 for Zero-Shot Antibody Design

Chai Discovery raised $130M at a $1.3B valuation in December 2025, backed by OpenAI and others, after unveiling Chai-2 with a 100x improvement in zero-shot de novo antibody design and a 16% hit rate.

·2 min

Stay current

Weekly digest of AI drug discovery developments. No noise.