On January 28, 2026, Google DeepMind released the source code and model weights for AlphaGenome, a unified DNA sequence model that predicts thousands of molecular properties from up to one million base pairs of DNA sequence at single-nucleotide resolution. The release, accompanied by a publication in Nature, makes one of the most powerful genomics AI models freely available for noncommercial research.
A Unifying Model for Genomics
AlphaGenome takes a fundamentally different approach from previous genomics models by unifying the prediction of multiple data modalities into a single framework. The model takes a long DNA sequence as input and predicts gene expression, transcription initiation, chromatin accessibility, histone modifications, transcription factor binding, chromatin contact maps, splice site usage, and splice junction coordinates and strength. This multi-task architecture allows the model to capture the complex interplay between different layers of gene regulation.
State-of-the-Art Performance
Trained on human and mouse genomes, AlphaGenome matches or exceeds the strongest available external models in 25 of 26 evaluations of variant effect prediction. The model's ability to process sequences up to one million base pairs long is critical for capturing long-range regulatory elements, such as enhancers that control gene expression from tens or hundreds of thousands of base pairs away. Its base-pair resolution enables the detection of fine-grained biological details that coarser models miss.
Rapid Research Adoption
Since AlphaGenome was first made available via API preview seven months before the open-source release, nearly 3,000 scientists from 160 countries have begun using it to advance research into cancer, neurodegenerative disorders, and infectious diseases. The open-source release of the full model code and weights is expected to dramatically accelerate adoption, enabling researchers to run the model on their own infrastructure, fine-tune it for specific applications, and build new tools on top of it.
Applications in Drug Discovery
AlphaGenome has immediate applications in drug discovery and precision medicine. By accurately predicting the functional impact of genetic variants, the model can help identify disease-causing mutations, prioritize drug targets based on genetic evidence, and predict the effects of therapeutic interventions at the genomic level. Researchers are already using it to investigate the functional significance of variants in the vast non-coding regions of the genome, sometimes called genomic 'dark matter,' which harbor regulatory elements critical to human disease but have been largely inaccessible to previous computational methods.
Broader Impact
The release follows the pattern established by AlphaFold, where open-sourcing a frontier AI model catalyzed a wave of scientific discovery. DeepMind's decision to make AlphaGenome freely available underscores a commitment to accelerating biological research through open science, while maintaining commercial applications through Isomorphic Labs. The model represents a significant step toward the long-term goal of computationally understanding the complete regulatory logic of the human genome.
