DeepMind Launches AlphaGenome 2 for DNA Variant Prediction

Google DeepMind released AlphaGenome 2 today, a DNA sequence model trained on more than 100 million individual cell profiles across 68 tissues of bulk and single-cell data from human and mouse samples. On the TenK10K dataset of 28 primary immune cell types, the model identified causal variants 30% of the time, against 16% for the previous approach, and got the direction of gene effects right 43% of the time versus 23%.

The pitch is adaptability rather than raw accuracy alone. "What makes AlphaGenome 2 exciting is how it adapts directly to new single-cell data without retraining, predicting how DNA variants act in newly profiled cell states from even a small sample of cells," said Sarah Teichmann, co-founder of the Human Cell Atlas and professor of stem cell medicine at Cambridge, in DeepMind's post. Peter Donnelly, CEO of Genomics Limited, called it "an impressive advance in the field."

DeepMind frames the problem as unfinished work: cells share identical DNA but switch different genes on and off, and predicting how subtle DNA variations ripple through that expression remains, in the company's words, an unsolved challenge and a key driver of human traits and disease. Doubling the share of causal variants a model can flag is the kind of jump that matters to anyone screening genomes for disease drivers.

Access is split. The model is available today for non-commercial research via the AlphaGenome API, with commercial research in private preview and AlphaGenome 2 Skills coming to Google Antigravity. The post names Adam Gayoso, Natasha Latysheva, Pushmeet Kohli and Žiga Avsec among the contributors.

It is the second DeepMind-flagged life-sciences release in a stretch that also saw Alphabet's Gemini for Healthcare launch with Northwell, HCA and Optum earlier the same day, part of a run of 80 healthcare AI stories we have tracked in the last 90 days.