AlphaGenome: advancing regulatory variant effect prediction with a unified DNA sequence model

Deep learning models that predict functional genomic measurements directly from DNA sequence are powerful tools for deciphering the genetic regulatory code. However, existing methods typically trade off input sequence length, prediction resolution, and modality coverage, limiting their scope and performance. AlphaGenome addresses these limitations through a unified framework that takes a 1-megabase DNA sequence as input and predicts thousands of functional genomic tracks at up to single-base-pair resolution across diverse modalities. Its outputs span 5,930 human and 1,128 mouse tracks, including gene expression, chromatin accessibility, histone modifications, DNA contact maps, and splice-junction coordinates and strengths. Trained on human and mouse genomes, AlphaGenome matches or exceeds the strongest available external models across a broad range of genome-track and variant-effect prediction benchmarks. Its performance extends across multiple regulatory mechanisms, including gene expression, splicing, and chromatin regulation. For gene expression, AlphaGenome substantially improves prediction of eQTL effect direction, recovering more than twice as many GTEx fine-mapped eQTLs as the previous leading model at a 90% sign-accuracy threshold. For splicing, a composite scorer integrating splice-junction predictions achieves state-of-the-art performance across multiple benchmarks, including the classification of pathogenic and benign ClinVar variants. The model also performs strongly in predicting chromatin accessibility QTL effects across diverse human ancestries. Importantly, AlphaGenome can simultaneously evaluate variant effects across multiple regulatory modalities, enabling the reconstruction of complex molecular mechanisms underlying clinically relevant variants. This is illustrated near the TAL1 oncogene, where the model recapitulates experimentally validated effects of gain-of-function mutations across gene expression, histone modifications, and transcription-factor binding. To facilitate broader use, AlphaGenome provides tools for making genome-track and variant-effect predictions directly from DNA sequence.