Resources#
Videos#
These videos were recorded for AlphaGenome 1. Core concepts such as intervals, variants, output modalities, variant scoring, and plotting carry over to AlphaGenome 2; see Migrating from AlphaGenome 1 for API differences.
Tutorial AlphaGenome 1
Walkthrough of AlphaGenome 1’s multimodal sequence predictions, variant effect scoring, and Python API workflow.
Roundtable AlphaGenome 1
Google DeepMind authors discuss the biological and engineering design behind AlphaGenome 1, the Nature paper, and the API.
Publications#
Paper Nature · 2026
Advancing regulatory variant effect prediction with AlphaGenome
Avsec et al. Introduces AlphaGenome 1, a unified 1 Mb DNA sequence model, and benchmarks multimodal predictions and variant effect scoring across gene expression, splicing, chromatin features, and contact maps.
Preprint medRxiv · 2026
AlphaGenome Atlas: in silico mutagenesis of the entire human genome improves prioritization and interpretation of non-coding variants
Cheng et al. Presents genome-wide in silico mutagenesis across 9 billion human variants and the AlphaGenome Variant Impact (AVI) score for variant prioritization.
See How should I cite AlphaGenome? for BibTeX citation entries.
Agent skills#
The Google DeepMind Science skills repository provides skills for Google Antigravity to run AlphaGenome and AlphaGenome Atlas workflows from natural-language prompts. To enable the Science plugin in Google Antigravity, open Customizations > Build with Google > Science > Add.
Agent skills Antigravity
Google DeepMind repository of scientific skills for Google Antigravity, including AlphaGenome skills for single-variant regulatory and splicing analysis via the API, variant prioritization and annotation with AlphaGenome Variant Impact (AVI) scores, and generating deep links to the AlphaGenome Atlas web application.
Model weights and code#
The alphagenome package queries the hosted API and requires no local GPU or
TPU hardware. If you want to download and run the AlphaGenome model weights
locally, see the repository below.
Model weights JAX
Google DeepMind research repository for using the AlphaGenome model weights, including the JAX reference implementation, variant scorers, and evaluation notebooks.
Data downloads#
To download precomputed AlphaGenome artifacts, visit the downloads page below.
Data
Precomputed AlphaGenome artifacts, including variant effect scores, AVI scores, and enhancer-gene links.