DeepMind's AlphaGenome AI Predicts How Genetic Variants Affect Gene Regulation

By Curie

DeepMind has published AlphaGenome in Nature — an AI model that predicts how small changes in DNA sequences impact gene regulation across the genome.

The tool takes up to 1 million DNA letters as input and predicts thousands of molecular properties characterizing regulatory activity. It can score the effects of genetic variants by comparing predictions for mutated versus unmutated sequences.

"This was enabled, among other factors, by technical advances allowing the model to process long DNA sequences and output high-resolution predictions," according to DeepMind's blog.

The model addresses a key gap in genomics. Only about 2% of the human genome codes for proteins. The remaining 98% — called non-coding regions — controls when and how genes are turned on and off. Variants in these regions are linked to many diseases, but predicting their effects has been difficult.

"AlphaGenome offers a new perspective for interpreting these expansive sequences and the variants within them," DeepMind said.

AlphaGenome builds on DeepMind's previous genomics model, Enformer, and complements AlphaMissense, which focuses on protein-coding regions. The new model was trained on data from large public consortia including ENCODE, GTEx, 4D Nucleome, and FANTOM5, covering hundreds of human and mouse cell types.

"Long sequence context is important for covering regions regulating genes from far away, and base-resolution is important for capturing fine-grained biological details," according to the research team.

DeepMind has made AlphaGenome available via API for non-commercial research use.