Google DeepMind Launches AlphaGenome 2 to Decipher DNA Mutation Impacts on Disease and Cell Types
Google DeepMind releases AlphaGenome 2, a breakthrough biological AI model that predicts how DNA sequence variants alter gene regulation in specific cell types, doubling disease-variant discovery rates.
Google DeepMind has officially launched AlphaGenome 2, an advanced biological foundation model capable of deciphering how non-coding DNA sequences and genetic variations regulate gene activity across distinct human cell types and disease states. Moving computational biology beyond protein 3D structure prediction into functional genomics, the model doubles the hit rate of previous systems in pinpointing causal disease-linked mutations.
Beyond Protein Folding: Functional Genomic Interpretation at Single-Cell Resolution
While AlphaFold revolutionized structural biology by predicting 3D protein structures from amino acid chains, over 98 percent of the human genome consists of non-coding DNA that orchestrates when, where, and how genes are transcribed. AlphaGenome 2 operates as a deep-learning sequence foundation model trained across massive genomic databases, mapping how single-nucleotide polymorphisms (SNPs) perturb chromatin accessibility, histone modifications, and transcription factor binding affinities within specific cell types and states.
Doubling the Causal Disease Variant Hit Rate Across Human Tissues
In empirical benchmarks reported by DeepMind, AlphaGenome 2 achieved a dramatic performance leap over its predecessor. On rigorous immune-cell validation challenges, the model successfully identified 30 percent of causal regulatory variants linked to autoimmune and oncological diseases, compared to just 16 percent for the first-generation system. Crucially, AlphaGenome 2 can adapt dynamically to incorporate newly mapped single-cell RNA sequencing data without requiring end-to-end retraining.
Research API Access, Antigravity Agent Integration, and Clinical Horizons
DeepMind has opened immediate access to AlphaGenome 2 for non-commercial academic researchers through a dedicated web API, with commercial enterprise evaluations entering private preview. The company also announced forthcoming AlphaGenome 2 Skills integration within Google Antigravity, its agentic biological development environment, enabling autonomous scientific agents to design targeted cellular therapies and prioritize drug candidates directly from patient sequencing data.