AlphaGenome — Google DeepMind

Using AI to decipher the genome
Understanding the genome is key to learning how biology works, why diseases occur and how to treat them. AI is a key tool to make this understanding possible.
The human genome consists of both the direct blueprints that code for proteins as well as the instructions that orchestrate and regulate gene activity telling the body when, where and how much of a certain protein to make.
Interpreting how genetic variations impact this complex system of gene expression and its effects at a molecular level has been a difficult experimental process. AlphaGenome can help accelerate progress toward that challenge. It is a set of advanced AI models and tools that can predict how genetic variants disrupt these biological processes, revealing how both the blueprints and the control panel are wired together.
It provides a comprehensive catalog of AlphaGenome predictions, and features the AlphaGenome Variant Impact (AVI) score, a unified metric that evaluates variants giving researchers a comprehensive measure to understand the molecular impact of changes in our DNA.
Deciphering how cells read DNA to better understand disease
Trained on more than 100 million individual cell profiles across 68 tissues, AlphaGenome 2 identifies nearly twice as many gene expression-linked variants in specific human cell types as our previous model.AlphaGenome 2 also generalizes to new experiments without fine-tuning. Scientists can upload their own single-cell RNA-seq data to adapt the model to unseen biological states and analyze cellular changes across disease progression, drug treatments, or early development.
By bringing single-cell precision to human genetics, AlphaGenome 2 gives scientists a powerful new tool to help uncover the cellular root causes of disease.
AlphaGenome Atlas is the most comprehensive prediction catalog of how single-letter variant changes in the human genome affect molecular biology.
We’ve used AlphaGenome to predict the molecular impact of every possible single-letter change in the human genome. All nine billion of them.
AlphaGenome Atlas is a 1PB dataset containing every single possible nucleotide variant alongside an AlphaGenome Variant Impact (AVI) score. This score helps researchers quickly prioritize between different variants across the whole genome, both those that produce proteins and control gene activity, while also interpreting their molecular effects.
AlphaGenome Atlas is available to researchers around the world, with coding and non-coding variations available.
Accelerating genomic research with Google Antigravity
The AlphaGenome and AlphaGenome Atlas Skills empowers researchers to run complex workflows directly within Google Antigravity—an AI-powered scientific workbench.
By connecting an AI assistant directly to the 1PB AlphaGenome Atlas dataset, researchers can transition from manual data searching to automated hypothesis generation in minutes.
Rank massive lists of uncharacterized genetic errors using the AlphaGenome Variant Impact (AVI) score.
Receive AI-generated breakdowns explaining why a variant is highly ranked, pinpointing specific biological disruptions like splicing defects.
Generate visualizations, such as ref/alt plots, directly inside your chat window without writing a single line of code.
The Atlas skill automatically generates parameterized URLs, allowing you to jump straight from your query directly to the relevant tracks on the Atlas website.
“The synergistic approach of using pre-computed AVI scores to prioritise variants from a starting point of millions, followed by visualising the predicted impact of specific variants in Atlas, is a powerful leveraging of both breadth and precision to identify novel interesting candidates for rare disease. As a non-coding specialist, there has never been the same array of tools available to us for in silico non-coding predictions as for protein-coding variants, so I see this as a really important step towards a time when we can recognise the contribution of these regions to rare disease.”
Laura CovillBroad Institute / GREGoR Consortium
“The human genome is a massive search space. Finding a genetic variant that matters for a specific disease or trait can be a difficult process, like finding a needle in a haystack. That’s what makes AlphaGenome Atlas so helpful - we can use it to shrink the haystack, maximize our chances of finding the needle, and ultimately make more scientific discoveries faster.”
Dr. Gareth HawkesUniversity of Exeter
“This is a very difficult problem as every cell type 'speaks a slightly different language' and thus it is hard to know what rules are general. With AlphaGenome, we can quickly query lots of cell types and look for general patterns by which genes are activated and repressed.”
Julia ZeitlingerStowers Institute for Medical Research
“What excited me about AlphaGenome is that we're no longer limited to understanding how disease risk variants affect gene expression alone. You can take a variant, mutate it in computationally, and understand its effects across the whole regulatory landscape - splicing, chromatin, even how the genome folds. For Alzheimer's, where most risk variants sit in the “dark matter” of the genome, outside of genes, that was huge. It's what allowed us to trace the genetic risk signals of Alzheimer’s out of the brain and into immune and metabolic tissues.”
César CunhaUniversity of Copenhagen
Accelerate your genomics research
A free, interactive portal for visual analysis with zero coding expertise required.
Accelerate your genomics research using AlphaGenome Skills on Antigravity.
Programmatic access to AlphaGenome 1, 2 and the Atlas dataset
Access AlphaGenome for commercial and enterprise workflows on Google Cloud via Model Garden on Gemini platform
I accept Google's Terms and Conditions and acknowledge that my information will be used in accordance with Google's Privacy Policy.