RELEASE · MODELS · #3431
DeepMind releases AlphaGenome 2 to predict cell-type‑specific effects of genetic variants
DeepMind unveiled AlphaGenome 2, a successor to AlphaGenome that predicts how DNA variants affect specific cell types and states. Trained on over 100 million single-cell profiles across 68 tissues, the model reportedly detects nearly twice as many expression-altering variants as the prior state of the art and is being made available for non-commercial research via the AlphaGenome API with a limited private preview for commercial use.
KEY POINTS
- DeepMind unveiled AlphaGenome 2, a successor to AlphaGenome that predicts how DNA variants affect specific cell types and states.
- Trained on over 100 million single-cell profiles across 68 tissues, the model reportedly detects nearly twice as many expression-altering variants as the prior state of the art and is being made available for non-commercial research via the AlphaGenome API with a limited private preview for commercial use.
- AlphaGenome 2 brings finer cell-type and state resolution to genetics, helping link population-scale variants to disease mechanisms and potential therapeutic targets.
DECISION BRIEF
CONFIDENCE
WHAT CHANGED
DeepMind released AlphaGenome 2, a successor to AlphaGenome that predicts the effects of genetic variants at cell-type and cell-state resolution. The model was trained on over 100 million (reported as >100M and >107M in sources) single‑cell profiles across 68 tissues, operates as a conditional model that takes a DNA sequence plus a specified biological state, and reports nearly twice as many expression‑altering variants versus the prior state of the art. It can predict gene activity, chromatin packaging/accessibility and DNA methylation, infer missing measurement modalities, and — according to reporting — is available for noncommercial research.
WHY NOW
AlphaGenome 2 claims finer, cell‑level resolution that can (per the sources) link population-scale genetic variants to cellular effects, help identify gaps in experimental data, and prioritize which lab measurements to run next — potentially accelerating researchers’ ability to connect variants to disease-relevant cell types and states.
WHO IS AFFECTED
Directly affected: genomics researchers and labs working with single‑cell datasets, researchers linking population genetics to cell biology, and groups planning experiments to test variant effects. The reporting references both human and mouse single‑cell profiles in training data.
CONFIRMED
Explicit, source-backed facts: 1) DeepMind has published a model called AlphaGenome 2 that predicts effects of genetic variants on individual cell types and states. 2) The model was trained on more than 100 million individual cell profiles (sources report >100M and >107M) across 68 tissues. 3) AlphaGenome 2 is described as identifying nearly twice as many variants that change gene expression compared to the previous state of the art. 4) The model operates conditionally (DNA sequence + specified biological state) and outputs gene activity, DNA packaging/accessibility, and DNA methylation predictions. 5) It can infer missing data modalities from single measurements. 6) Reports state it is available for noncommercial research.
UNCERTAIN
Missing or conflicting evidence (explicitly labeled): 1) Commercial access terms and whether a limited private preview for commercial use exists are not confirmed in the supplied excerpts (one excerpt explicitly states noncommercial availability; no excerpt details commercial preview or API terms). 2) Exact training dataset composition and provenance (which datasets, how distributed between human vs. mouse, preprocessing) are not fully specified in the excerpts. 3) Detailed benchmark numbers, evaluation metrics, and peer‑reviewed validation beyond the company‑reported ‘nearly twice as many’ are not provided in the supplied excerpts (a benchmark name is partially referenced but results are not shown). 4) Downstream impact on specific disease discoveries or therapeutic targets is suggested by sources but not demonstrated with explicit results in the excerpts. 5) Any limitations, failure modes, or reproducibility checks are not described in the supplied excerpts.
WHAT TO WATCH
Concrete observable signals to watch next: 1) DeepMind blog post, technical preprint, or peer‑reviewed paper releasing full methods, datasets, and metrics. 2) Public benchmark results and detailed evaluations (e.g., full TenK10K benchmark output referenced in reporting). 3) Documentation or portal announcements detailing access modalities, API, and any commercial preview terms. 4) Independent reproduction studies or community benchmarks comparing AlphaGenome 2 to other models. 5) Reports from labs using the model to nominate experiments, identified gaps in experimental data, or to link variants to specific disease biology.
WHY IT MATTERS
AlphaGenome 2 brings finer cell-type and state resolution to genetics, helping link population-scale variants to disease mechanisms and potential therapeutic targets.
EVIDENCE MAP
4Editorial claims linked to specific sources, with support, contradiction and context shown separately.
DeepMind released AlphaGenome 2, an AI model that predicts how genetic variants affect specific cell types and cell states.
SUPPORTEDChecked 2026-11-12 · 2 supporting
AlphaGenome 2 was trained on over 100 million single‑cell profiles spanning 68 tissues.
SUPPORTEDChecked 2026-11-12 · 2 supporting
The Verge reports the model was trained on "more than 100 million individual cell profiles across 68 tissues."
SUPPORTS The DecoderMEDIA · 2026-11-12The Decoder reports training on "more than 107 million individual cell profiles from 68 tissues" (notes data include humans and mice).
In evaluations reported by DeepMind, AlphaGenome 2 identifies nearly twice as many variants that alter gene expression compared with the prior state of the art / its predecessor.
SUPPORTEDChecked 2026-11-12 · 2 supporting
The Verge says DeepMind reports AlphaGenome 2 "identifies nearly twice as many variants that can change gene expression" versus the previous state of the art.
SUPPORTS The DecoderMEDIA · 2026-11-12The Decoder similarly states the model "found nearly twice as many relevant gene changes as its predecessor."
AlphaGenome 2 can generalize to new cell types and biological states without retraining and can infer missing experimental measurements.
SUPPORTEDChecked 2026-11-12 · 2 supporting
The Verge excerpt reports the model "could make predictions about new types of cells and biological states without needing to be retrained" and that it "could also identify gaps in experimental data."
SUPPORTS The DecoderMEDIA · 2026-11-12The Decoder notes the model "can infer missing data from a single measurement" (providing predictions for additional properties when only one is measured).
SOURCES & TIMELINE
2DeepMind’s new DNA model better predicts how mutations affect different cells, potentially opening paths to treatment. Researchers at Google DeepMind have unveiled a new AI model designed to unravel how changes in our DNA affect different cells, giving scientists a more detailed view of a notoriously difficult problem that could hold important clues to the causes of disease — and point the way to new treatments. Al…
Google Deepmind has introduced AlphaGenome 2, an AI model that predicts gene changes in specific human cell types. The system analyzes individual cell states and computes gene activity, physical DNA packaging, and chemical switches on the DNA. In testing, the model found nearly twice as many relevant gene variants as its predecessor and is now available for noncommercial research. Google Deepmind has released Alph…