AlphaGenome Atlas Scores Genetic Changes for Disease Research
Google DeepMind said in a blog post on Nov. 17 that its AlphaGenome tool and a companion dataset, AlphaGenome Atlas, are helping researchers search human DNA for changes that may cause disease. The post is an interview with Žiga Avsec, a DeepMind research scientist who leads its genomics work.
The human genome has 3 billion base pairs. Each one can swap to three other letters, which makes 9 billion possible single-letter changes. Avsec said DeepMind used AlphaGenome to predict the effects of all of them. Atlas turns those predictions into a score, called AVI, for each change. Researchers can use it to rank which variants look most disruptive.
That ranking matters because the haystack is huge. Avsec described a patient with a rare disease whose DNA might differ from the reference genome in 4 million places. Even after filtering for rare ones, hundreds or thousands can remain. Lab experiments can sort them out, but he said they are costly and slow.
Two outside researchers are already using it, according to the post. Gareth Hawkes of the University of Exeter narrowed 1 billion variants to the top 1% and found 19 variant groups tied to body mass index. Laura Covill of the Broad Institute used the score on unsolved rare diseases. Atlas flagged a variant in a gene linked to neurodevelopmental disorders, which helped solve the case.
For most people, any benefit is indirect. Avsec said the tools could help diagnose rare genetic disorders, personalize treatment and speed up drug discovery. Other labs are also building biology systems, such as OpenAI's biology model.
Several things are still open. DeepMind says it is working on better ways to show how confident the model is in each prediction. Avsec said people should read the results alongside other sources of information. A newer version, AlphaGenome 2, aims to predict effects in individual cell types, since the first model was trained mostly on whole-tissue or cell-line data. The post gives no results from that version. A full digital simulation of a human cell remains a long-term goal. More systems like this one are in our list of AI models.
Sources
Cite this
AlphaGenome Atlas Scores Genetic Changes for Disease Research, New Mind Museum, November 17, 2026. https://newmindmuseum.com/research/2026/11/alphagenome-atlas-disease-research/
New Mind Museum. (2026, November 17). AlphaGenome Atlas Scores Genetic Changes for Disease Research. https://newmindmuseum.com/research/2026/11/alphagenome-atlas-disease-research/
“AlphaGenome Atlas Scores Genetic Changes for Disease Research.” New Mind Museum, 17 Nov. 2026, https://newmindmuseum.com/research/2026/11/alphagenome-atlas-disease-research/.
@misc{new-mind-museum-2026-alphagenome-atlas-scores-genetic-changes,
author = {{New Mind Museum}},
title = {AlphaGenome Atlas Scores Genetic Changes for Disease Research},
year = {2026},
url = {https://newmindmuseum.com/research/2026/11/alphagenome-atlas-disease-research/}
}Free to quote and cite with a link to this page.