What researchers are trying
AlphaGenome processes up to one million DNA bases at single-base resolution and predicts thousands of regulatory tracks and variant effects.
Million-base regulatory variant prediction
Can a single model predict how coding and non-coding variants alter expression, splicing, chromatin and regulatory binding over long genomic context?
Why it matters
Most disease-associated variants are non-coding. Better regulatory predictions could connect association signals to mechanisms, targets and experiments.
Evidence boundary
The model is a research predictor, not a personal-genome or clinical diagnostic system; tissue specificity and very long-range enhancer logic remain limitations.
Organizations represented
Google DeepMind
Demonstrated evidence
What has actually been shown.
- Reported strong performance across sequence and variant-effect benchmarks and mechanistic case studies.
Unresolved questions
- How reliably do predictions transfer to rare cell states and patient contexts?
- Can causal mechanisms be separated from learned correlations?
- How should predictions be prospectively validated?
Signals to watch next
- Independent variant benchmarks
- Disease-specific validation
- Cell-state conditioning
- Clinical governance
Connected evidence graph
Related BioAtlas model passports.
AlphaGenome
Reading the genome's 'dark matter' at base-pair resolution.
2/7 evidence fields documentedEnformer
Transformers that predict gene expression from raw sequence.
2/7 evidence fields documentedAlphaMissense
Classifying which missense mutations cause disease.
2/7 evidence fields documentedNucleotide Transformer
BERT-style foundation models for the human genome.
3/7 evidence fields documentedPrimary and evaluation sources