AlphaGenome AI model may help decode how DNA influences disease

Google DeepMind's AlphaGenome is a sequence-to-function AI that can analyse up to one million letters of DNA and predict how genetic variants affect gene regulation and expression. The model has been described in Nature and made available for non-commercial research, and roughly 3,000 researchers have used it.

AlphaGenome is an AI tool from Google DeepMind that looks at DNA sequences to predict how changes in the code affect biological function. It is a sequence-to-function model trained on public human and mouse cell experiment databases and was reported in the journal Nature. The system can analyse long stretches of DNA at once and predict effects on gene activity, splicing and other regulatory signals. DeepMind made the model available for non-commercial research in 2025 and several thousand scientists have used it. Key points: - AlphaGenome was trained on publicly available human and mouse cell experiment datasets and is described in Nature. - The model can analyse up to about one million letters of DNA at a time and predicts gene location, expression levels, splicing, and the effect of single-letter changes. - DeepMind reports the tool has been used by roughly 3,000 researchers via an API since mid-2025, and researchers worldwide have applied it to studies of cancer, metabolic conditions and other diseases. - Independent groups report strong performance in many tests, with noted limitations in predicting long-range gene regulation (over ~100,000 letters) and in some tissue-specific predictions. Summary: Researchers say AlphaGenome could help prioritise genetic variants for laboratory study and inform work on disease mechanisms, drug targets and synthetic biology design. The developers and outside experts also report that the model needs further refinement, better training data and improved accuracy across tissues and long-range regulation, and a commercial version is under early testing. Undetermined at this time.