Google DeepMind Launches AlphaGenome AI to Decode Human Genome, Published in Nature

Google (GOOGL-US) DeepMind unveiled AlphaGenome, a breakthrough AI model that predicts 11 gene regulatory processes simultaneously — from transcription factor binding to 3D chromatin folding — marking a leap beyond protein structure prediction into full-genome functional mapping. Published on the cover of Nature (January 29, 2026), it outperformed existing models in 22 of 24 genomic benchmarks. AlphaGenome excels in predicting effects of non-coding DNA mutations, resolving 50% of GWAS-linked variants by identifying distant enhancers up to 10kb away. In a leukemia case study, it traced an 8,000-base-pair-distant insertion that created a rogue enhancer activating the TAL1 oncogene. Built on a CNN-Transformer hybrid architecture and trained on human/mouse genomic data, it enables precision medicine by modeling RNA splicing errors and rare disease mechanisms.