Google’s AlphaGenome predicts DNA sequence function

In 2024, the creators of the AlphaFold machine learning model for predicting protein structure were awarded the Nobel Prize in Chemistry. And now DeepMind—Google’s artificial intelligence research laboratory where AlphaFold was developed—has officially released a new machine learning model called AlphaGenome that can predict the function of DNA sequences (Nature 2026, DOI: 10.1038/s41586-025-10014-0).AlphaGenome has been publicly available for noncommercial use since June 2025, when a preprint describing the model was released, and the latest version is available on GitHub (bioRxiv 2025, DOI:10.1101/2025.06.25.661532). Pushmeet Kohli, vice president of research at Google DeepMind, calls AlphaGenome “our solution to deciphering the complex regulatory code” that makes up most of a genome.In 2023, the team that worked on AlphaGenome released AlphaMissense, which can predict the effect of mutations in the protein-coding regions of the genome (Science 2023, DOI: 10.1126/science.adg7492). But AlphaGenome takes DNA functional prediction to a new level. Žiga Avsec, a Google DeepMind researcher and lead author of the AlphaGenome paper, says that the new model can predict “gene expression, DNA accessibility, histone modifications, transcription factor binding, and even folding structure of the genome,” with a high level of accuracy in sequences of DNA up to a million base pairs long.The researchers at Google DeepMind have sought feedback from scientists about the utility of their early-release versions of AlphaGenome, with Kohli saying it has allowed researchers to pinpoint the mutations in cancer genomes that drive proliferation of the cancer, among other uses. But only time will tell how much impact AlphaGenome will have