DeepMind announced AlphaGenome 2, a cell-type DNA model
Google DeepMind announced AlphaGenome 2, a DNA-sequence model it says can predict how genetic sequences and their variations affect specific human cell types. The company blog post is dated November 11, 2026, with a publication timestamp of 3:00 p.m. UTC, which is 10:00 AM ET. This article summarizes that announcement, a technical paper DeepMind linked from the same post, and contemporaneous coverage. It is not a laboratory evaluation of the model, and it is not medical advice.
Real, reported: based on the public sources cited below. Nothing added or dramatized. Sources checked November 12, 2026, at 7:18 AM ET (EST). The model's design, training data, and performance figures are company and laboratory claims. The opened blog post and the first pages of the linked paper do not name a scientific journal and do not describe the work as peer-reviewed. Those claims should be read as a company claim, not as an independently verified or peer-reviewed result.
What DeepMind says the model does
DeepMind wrote: "Today, we are announcing AlphaGenome 2, our next-generation model that predicts how DNA sequences and their variations impact specific human cell types, giving scientists a powerful new tool to uncover the cellular drivers of disease." The post says last year's AlphaGenome model predicted how DNA variants affect molecular activity across broad tissues, and that the lab "expanded its reach with AlphaGenome Atlas, a platform containing predictions for all 9 billion possible single-letter changes in the human genome." AlphaGenome 2, DeepMind says, zooms in from those broad tissue averages to individual cell types and cell states.
On performance, the blog states: "AlphaGenome 2 identifies nearly twice as many variants that change gene expression in cells, compared to our previous state-of-the-art model." It also says the model was "trained on more than 100 million individual cell profiles across 68 tissues" and "is available today for non-commercial research via the AlphaGenome API, and in private preview for commercial research." Those comparisons and counts are DeepMind's.
The same post is explicit about intended use: "Please note that AlphaGenome and AlphaGenome 2 are intended for research use only and have not been validated or approved for clinical or diagnostic purposes."
Later in the blog, DeepMind gives a more specific training figure, saying it trained the model "on both bulk and single-cell data from human and mouse samples, including over 107 million individual cell profiles." It describes AlphaGenome 2 as a conditional model: the same stretch of DNA can be read differently across cell types, and, DeepMind says, if a scientist inputs one measurement such as single-cell gene activity, the model can then "predict unmeasured properties, such as chromatin accessibility." The post says the model predicts the same core readouts as the earlier AlphaGenome, including gene expression, histone modifications and chromatin accessibility, and that it also predicts DNA methylation.
On a dataset DeepMind calls TenK10K, "a large single-cell map of 28 primary immune cell types," the blog says AlphaGenome 2 "reliably identified nearly twice as many causal variants (30% vs. 16%) and directional gene effects (43% vs. 23%) as the original AlphaGenome fine-tuned on single-cell data and other specialized tools." Those percentages are DeepMind's reported comparison, not a result we reproduced. The post also says the model computationally reproduced a published finding in which a distant regulatory variant drives inflammatory bowel disease by overactivating the ETS2 gene in gut macrophages, singling out the causal variant and leaving other cell types and genes largely unaffected, according to DeepMind.
The same announcement includes a comment from Professor Sarah Teichmann, identified on the blog as a co-founder of the Human Cell Atlas consortium and a professor of stem cell medicine at the University of Cambridge: "What makes AlphaGenome 2 exciting is how it adapts directly to new single-cell data without retraining, predicting how DNA variants act in newly profiled cell states from even a small sample of cells." That comment appears on DeepMind's announcement page. It is not an independent paper by Teichmann, and it is not a clinical endorsement.
What the linked technical paper adds
The blog includes a "Read the paper" button that points to a PDF hosted by DeepMind. The opened pages identify the authors as being at Google DeepMind and present AlphaGenome 2 as a model that takes a 1-Mb DNA sequence plus a condition specifying a target tissue, cell type, or cell state. The paper says the model "outperforms fine-tuned AlphaGenome 1 and specialized models trained on target datasets, doubling recall of causal single-cell eQTLs at 90% precision." It also says the authors generated an atlas of enhancer-gene links and scores for 7 million common variants "across hundreds of human cell types." Those statements are from the laboratory paper DeepMind published with the announcement. We have not independently reproduced the measurements, and the opened pages do not present them as a journal publication.
How SiliconANGLE reported the release
SiliconANGLE, in an article first posted at 11:30 AM ET on November 11, reported that DeepMind announced AlphaGenome 2 "today," describing it as a next-generation AI model that can predict how DNA sequences and their variations affect human cell types. The outlet said these predictions are meant to guide research rather than replace experimental validation, and that the model "is not approved for clinical or diagnostic use." That matches DeepMind's own research-only warning. SiliconANGLE's piece also covers a separate DeepMind robotics announcement; the genomics claims above are the ones relevant to this article.
Our advice (general guidance)
Treat AlphaGenome 2 as a newly announced research tool from the lab that built it, not as a diagnostic product. DeepMind itself says the model has not been validated or approved for clinical or diagnostic purposes. Do not rely on its variant predictions to make medical, insurance, or treatment decisions. If you are a researcher trying the API or a preview, read the access terms on DeepMind's official pages, which you should open on your own, and keep private genomic or patient data off any system you have not reviewed.
Be cautious of third-party sites or messages that claim to offer "official AlphaGenome 2 access," a paid interpretation of your DNA, or a free clinical report. Those offers are a common setting for phishing and junk medical claims. Use the project pages DeepMind publishes, and ignore anyone who asks you to disable security tools or to send genetic files to an unfamiliar address.
These are editorial precautions, not individualized medical, scientific or legal advice.
Sources
- Google DeepMind: AlphaGenome 2: Deciphering how cells read DNA to better understand disease (Nov 11, 2026, 10:00 AM ET)
- Google DeepMind technical paper PDF linked from that blog post as 'Read the paper' (AlphaGenome 2 predicts regulatory activity and variant effects across cell states)
- SiliconANGLE: Google DeepMind releases new advanced robotics AI model and next-generation gene prediction model (first posted Nov 11, 2026, 11:30 AM ET; page also shows a later update)
Published November 12, 2026 by FraudMeNot Team. Label: Real, reported.