Google DeepMind announced AlphaGenome 2 on November 11, 2026, introducing a next-generation DNA sequence model that predicts how DNA sequences and their variations affect specific human cell types. According to the company, the model gives scientists a powerful new tool for uncovering the cellular drivers of disease, bridging single-cell biology and human genetics by forecasting how genetic variants influence hundreds of human cell types.

The announcement, authored by DeepMind researchers Adam Gayoso, Natasha Latysheva, Pushmeet Kohli and Žiga Avsec, is available on the Google DeepMind blog, accompanied by a technical report (PDF).

From a shared DNA blueprint to cell-specific gene activity

Every cell in the human 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 DeepMind notes that predicting how subtle DNA variations affect expression inside cells remains an unsolved challenge, and a key driver of human traits and disease.

Last year, the company launched AlphaGenome, a model predicting how DNA variants affect molecular activity across broad tissues, with the underlying research published in Nature. It later expanded the tool with AlphaGenome Atlas, a platform containing predictions for all 9 billion possible single-letter changes in the human genome. According to DeepMind, these tools help researchers prioritize and interpret variant effects, driving discoveries in rare diseases, complex traits and fundamental biology.

By zooming in from broad tissue averages to individual cell types and cell states, and redesigning the model's architecture, AlphaGenome 2 represents what the company calls a step-change improvement in variant predictive performance. It allows researchers to predict variant effects in disease-relevant cells that would be missed in broad-tissue models, as well as in cell and disease states the model has not seen before. DeepMind reports that AlphaGenome 2 identifies nearly twice as many variants that change gene expression in cells compared with its previous state-of-the-art model. The model was trained on more than 100 million individual cell profiles across 68 tissues, turning static cell atlases into what the company describes as a predictive engine for human genetics.

AlphaGenome 2 is available as of the announcement for non-commercial research via the AlphaGenome API, and in private preview for commercial research.

A conditional architecture built for specific cell types

The first AlphaGenome took a 1-million-letter DNA sequence and predicted properties across thousands of experiments from many tissues all at once. While efficient, DeepMind says, this parallel approach spread the model's capacity thin and made it hard to capture subtle differences between closely related cell types.

AlphaGenome 2 fundamentally redesigns how cell types are handled. Instead of predicting in parallel, it is a conditional model that predicts DNA effects for one condition at a time — with a condition defined as a tissue, cell type or state. Alongside the million-letter DNA sequence, the model takes the target biological condition as input at every stage of its architecture. The condition can be chosen from a reference catalog of cell types or generated on the fly from a scientist's own single-cell experiment via DeepMind's cell encoder. This design lets the model learn how the same stretch of DNA is read differently across cell types.

The conditional approach also enables cross-modality imputation — predicting missing experimental data. Because the model separates a cell's biological state from its outputs, a scientist who inputs a single measurement, such as single-cell gene activity, can obtain predictions for unmeasured properties such as chromatin accessibility. The company says this can help researchers prioritize which experiments to do next.

The model was trained on both bulk and single-cell data from human and mouse samples, including over 107 million individual cell profiles. Moving beyond the bulk tissue averages of the original model, AlphaGenome 2 can resolve activity inside specialized cells — such as distinct heart muscle or blood immune cells — to help uncover crucial DNA variants that act only in specific cell types. It predicts the same core experimental readouts of DNA function as the original AlphaGenome — gene expression, histone modifications and chromatin accessibility — while adding DNA methylation, a readout the company describes as widely requested.

A standout capability, according to DeepMind, is that the model 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.

State-of-the-art cell-type predictions and variant recall

DeepMind reports that AlphaGenome 2 is significantly better at predicting gene regulation in specific cells, helping researchers pinpoint disease-relevant variants.

Improved cell-type specificity. On genome regions outside its training data, the model explains significantly more of how gene expression differs across cell types within a tissue, tracking continuous changes in gene expression — such as when human blood stem cells mature into red blood cells and immune lineages. The announcement illustrates this with a heatmap of gene expression for maturing red blood cells (hematopoiesis), in which AlphaGenome 2's predictions are closer to row-normalized, Z-scored observations than those of the first model fine-tuned on single-cell data.

Improved variant effect prediction. On the TenK10K dataset, a large single-cell map of 28 primary immune cell types, AlphaGenome 2 reliably identified nearly twice as many causal variants (30% versus 16%) and directional gene effects (43% versus 23%) as the original AlphaGenome fine-tuned on single-cell data and other specialized tools. This precision, DeepMind says, helps researchers find overlooked candidate variants for new discoveries.

Pinpointing disease-relevant variants. A recent landmark study showed that a variant in a distant regulatory switch drives inflammatory bowel disease (IBD) by overactivating the ETS2 gene in gut immune cells known as macrophages. Making predictions across various immune cell types in the gut, AlphaGenome 2 reproduced this finding computationally: it singled out the causal variant and showed that it selectively boosts ETS2 in gut macrophages while leaving other cell types and genes largely unaffected.

Analyzing new cell types and states without retraining

AlphaGenome 2 can generalize to entirely new single-cell experiments without retraining. Scientists can upload a single-cell RNA-seq snapshot of gene activity from just a few dozen cells; this gene activity is compressed into a cell state representation that conditions the model on unseen states. According to DeepMind, AlphaGenome 2 can match or outperform models trained or fine-tuned directly on those datasets.

This enables the model to predict variant effects and gene regulation for new cell states, such as those arising from disease or vaccination. In one example, given unseen single-cell data from a COVID-19 mRNA vaccine study, AlphaGenome 2 distinguished vaccine-activated immune T cells from resting ones. It also revealed that a variant linked to severe COVID-19 disrupts gene activity more strongly when cells are in this activated state — an effect that would be missed in unactivated cells.

Bridging human genetics and single-cell biology

DeepMind frames better disease understanding as a link between human genetics and cell biology, requiring connections between the thousands of trait- and disease-associated variants found in genome-wide association studies (GWAS) and the growing single-cell maps of the human body. To help researchers make these connections without running large-scale computations themselves, the company precomputed AlphaGenome 2 predictions for trait-associated variants across hundreds of human cell and tissue conditions, tackling two major challenges:

• Mapping enhancer–gene links in real human cell types. Many disease-linked variants sit in distant regulatory switches called enhancers, making their effects hard to trace. Because experimental CRISPR screens mostly use stable, lab-grown cell lines, scientists often lack wiring diagrams for specific cells in the human body. AlphaGenome 2 extends enhancer–gene mapping to primary human cell types — such as brain astrocytes, which support and protect neurons — helping researchers connect regulatory variants to their target genes in the specific cells where disease unfolds.
• Pinpointing causal GWAS variants and the cells where they act. GWAS flag broad stretches of DNA linked to disease, but researchers cannot always identify the exact causal variant, target gene and cell type with high confidence. Across a number of complex traits, ranging from height to type 2 diabetes, AlphaGenome 2's precomputed variant scores help pinpoint the specific variants and cell types most likely to be involved. When confident, the model also accurately predicts whether a variant turns a gene's activity up or down — a critical clue, the company says, for understanding how that gene contributes to disease.

The blog post also included this statement, reflecting why drug-discovery teams are watching DNA sequence prediction models: "Human genetics is one of the most powerful tools in drug discovery. GWAS is central to that but it is difficult to move from an association signal to what a drug programme needs. DNA sequence prediction models are an important tool in helping bridge that gap, and AlphaGenome 2 represents an impressive advance in the field. We look forward to incorporating it in our Mystra platform."

Looking ahead with the scientific community

As scientists map more of the human body at the single-cell level, DeepMind expects AI models to continue becoming more powerful. Through ongoing collaborations, including with researchers at the Wellcome Sanger Institute, the company is working to profile an even wider spectrum of human cell types, developmental lineages and perturbed cellular states.

AlphaGenome 2, along with interactive Colab tutorials for expression conditioning, is available for non-commercial research via the AlphaGenome API and in private preview for commercial research. DeepMind also said it will soon release AlphaGenome 2 Skills in Google Antigravity, its agentic development platform, where specialized models like AlphaGenome 2 combined with agentic science tools such as Gemini 4 are intended to help scientists accelerate research and make new discoveries.

The company thanked its community of trusted testers and the many researchers who helped create the training datasets. It invited the community to submit interesting AlphaGenome 2 use cases via a form by November 30; selected projects may be eligible for additional quota and credits. Feedback to shape the future roadmap can be shared in the community forum.

The company noted that AlphaGenome and AlphaGenome 2 are intended for research use only and have not been validated or approved for clinical or diagnostic purposes, meaning their predictions are meant to guide scientific research rather than medical decisions.

Acknowledgements

According to the blog post, the core research work was done thanks to contributions from Adam Gayoso, Anne Mottram, Alexander Karollus, Guido Novati, Pavol Drotár, John Reid, Laetitia Meng-Papaxanthos, Adam Kosiorek, Eirini Arvaniti, Vincent Dutordoir, Soham De, Tom Ward, Lai Hong Wong, Matteo Perino, Kyle Taylor, Jun Cheng, Natasha Latysheva, Joshua Pan, Dhavanthi Hariharan, Clare Bycroft, Lauren Nicolaisen, Demis Hassabis, Pushmeet Kohli and Žiga Avsec.

DeepMind also thanked the following people for support and feedback: Gabriella Stanton, Charlie Taylor, Raina Thomas, Lucas Tenório, Antonio García, Joseph Powell, Angli Xue, Trygve Bakken, Nelson Johansen, Junyue Cao, Ziyu Lu, Jacob Hepkema, Leopold Parts, Irina Voineagu, Souradeep Basu, Michal Zielinski, Jeff Cao, Ioannis Sarropoulos, Sarah Teichmann, Amanda Stafford and Eliseo Papa. It further thanked Evie Gray, Adriana Fernandez Lara, Alex Wilkins, Danielle Breen, Dimple Vijaykumar, Muinat Abdul, Chris Kwaku-Abdullahi, Kathryn Seager, Ryan Smith, Ross West and Gaby Pearl for their expertise in communicating the work.