DeepMind Maps Nine Billion DNA Variants in AlphaGenome Atlas

Google DeepMind released AlphaGenome Atlas, a predictive map evaluating nine billion single-letter DNA mutations across the human genome. The resource calculates how individual genomic variants alter gene regulation and protein production, offering a unified database for functional genomics. As of March 2025, the platform gives biomedical researchers direct access to pre-calculated effect scores across hundreds of human cell types.

What's new

Google DeepMind built AlphaGenome Atlas to predict the molecular consequences of non-coding and coding DNA variants at scale. The database covers nearly nine billion potential single-nucleotide variants, evaluating their specific impact on RNA splicing, transcription factor binding, and epigenetic marks.

Rather than relying strictly on laboratory sequencing for every rare mutation, AlphaGenome Atlas uses deep neural networks to score variants based on predicted functional disruption. Researchers can input genomic coordinates to examine how a single base swap affects downstream cellular mechanisms. The complete dataset is available through an interactive web interface and downloadable files for non-commercial research.

How it fits your workflow

While produced by Google DeepMind—the research lab responsible for media models such as Google Veo 3—AlphaGenome Atlas serves computational biologists, geneticists, and clinical researchers rather than video editors or animators. It delivers regulatory predictions that speed up target discovery in disease research and therapeutic design.

In bioinformatics pipelines, AlphaGenome Atlas acts as a filter for high-throughput sequencing data, performing a role similar to tools like SpliceAI or Enformer. Instead of running separate predictive models for distinct regulatory marks, computational teams receive a consolidated profile across multiple biological dimensions. For clinical teams analyzing rare patient variants, this pre-computed index reduces the computational overhead required to select candidate mutations for laboratory testing.

What it costs / how to try it

AlphaGenome Atlas is accessible for free to academic and non-commercial researchers through Google DeepMind's web platform and public data repositories.