AlphaGenome 2

AlphaGenome 2 is the next-generation DNA sequence model Google DeepMind announced on 11 November 2026. The official write-up is AlphaGenome 2: Deciphering how cells read DNA to better understand disease, by Adam Gayoso, Natasha Latysheva, Pushmeet Kohli and Žiga Avsec. The official account @GoogleDeepMind posted at 2026-11-11T15:03:17Z, calling it “our new AI model that zooms in on specific cells to help scientists understand what drives disease.”

The blog opens with a basic fact of biology: every cell in the body carries the same DNA blueprint, yet each cell type and state — a heart muscle cell, or an immune cell fighting an infection — switches different genes on or off to do its job. Global efforts such as the Human Cell Atlas are mapping this gene expression for millions of cells, but predicting how subtle DNA variations affect that expression in specific cells remains an unsolved challenge, and a key driver of human traits and disease. Last year DeepMind launched AlphaGenome to predict how DNA variants affect molecular activity across broad tissues, then expanded its reach with AlphaGenome Atlas, a platform of predictions for all 9 billion possible single-letter changes in the human genome. AlphaGenome 2 zooms in from those broad tissue averages to individual cell types and cell states, and redesigned the architecture. The company says this is a step-change in variant predictive performance, and that the model can also make predictions for cell and disease states it has not seen before.

On training and capabilities, the official blog says the model was trained on both bulk and single-cell data from human and mouse samples, including over 107 million individual cell profiles across 68 tissues. Instead of predicting many tissues in parallel, AlphaGenome 2 is a conditional model: it predicts DNA effects for one condition — a tissue, cell type or state — at a time, and uses that target biological condition as input at every stage of its architecture, so it can learn how the same stretch of DNA is read differently across cell types. Besides the core readouts of the original model — gene expression, histone modifications and chromatin accessibility — it also predicts DNA methylation. A conditional design also allows cross-modality imputation: if a scientist inputs just one measurement, such as single-cell gene activity, the model can predict unmeasured properties such as chromatin accessibility. Scientists can also upload a single-cell RNA-seq snapshot from just a few dozen cells; that activity is compressed into a cell-state representation, and the model can then be conditioned on unseen states without retraining. Official examples include: on the TenK10K map of 28 primary immune cell types, 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 specialised tools; it also computationally reproduced a landmark finding that a variant in a distant regulatory switch drives IBD by overactivating ETS2 in gut macrophages, leaving other cell types and genes largely unaffected. Given unseen single-cell data from a COVID-19 mRNA vaccine study, it distinguished vaccine-activated immune T cells from resting ones and showed that a variant linked to severe COVID-19 disrupts gene activity more strongly in the activated state.

On availability, the blog says AlphaGenome 2 and interactive Colab tutorials for expression conditioning are available today for non-commercial research via the AlphaGenome API, and in private preview for commercial research. DeepMind also says it will soon release AlphaGenome 2 Skills in Google Antigravity. The post is explicit: AlphaGenome and AlphaGenome 2 are intended for research use only and have not been validated or approved for clinical or diagnostic purposes.

Key links:

• Official blog: https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/ ;
• Official post: https://x.com/GoogleDeepMind/status/2120532824465805749 (lead post at 2026-11-11T15:03:17Z, follow-ups at 15:03:18Z and 15:03:19Z);
• This is a specific model name from Google DeepMind, the next generation after the 2025 AlphaGenome. Do not merge it with AlphaGenome, AlphaGenome Atlas, or unrelated genomics models.

Why it matters

This is a formal model launch on 11 November 2026, with an official blog post and an official account thread in the same window — not a rumour or an internal alias. DeepMind is naming a next-generation, cell-type-conditional DNA model and opening it today for non-commercial research through the AlphaGenome API. For anyone tracking foundation models in genetics, that is a new, citable name.

This page only restates public official materials. Figures such as “nearly twice as many” expression-changing variants, the TenK10K 30% vs. 16% and 43% vs. 23% comparisons, and the 107 million cell profiles across 68 tissues are company statements and are not independently reproduced here. The official materials also state that the model is for research, not clinical or diagnostic use.

Limits

The public materials do not include a full benchmark table, an independent third-party reproduction, the terms of the commercial preview, or API quota details. The broader coverage of the launch notes that even at high precision the model recovers only a minority of the expression-linked and disease-associated variants that statistical analyses nominate as likely causal, and that it can underestimate effects outside accessible chromatin, at very distant regulatory elements, or through post-transcriptional mechanisms. Transient, stimulated and disease states remain underrepresented in training. Experimental follow-up is still required. Do not read this entry as medical or diagnostic advice.

1. Read the official blog https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/ and cross-check the announcement at https://x.com/GoogleDeepMind/status/2120532824465805749 and the follow-up posts in the same thread at 15:03:18Z and 15:03:19Z ;
2. For non-commercial research, use the AlphaGenome API and the interactive Colab tutorials as the official post describes; commercial research goes through the stated private preview, and the official blog says the models are for research only and have not been validated for clinical or diagnostic use ;
3. When citing “nearly twice as many causal variants,” “100 million+ cell profiles,” the conditional architecture, or the forthcoming AlphaGenome 2 Skills, say the source is the 11 November 2026 official blog or announcement, and re-check the upstream source ;
4. Do not treat AlphaGenome 2 as the 2025 AlphaGenome or as AlphaGenome Atlas, and do not use this page as clinical or diagnostic guidance; this page only restates the official blog and announcement — re-check the upstream source before citing it.