AlphaGenome

Google DeepMind’s genomic foundation model. Predicts chromatin accessibility, gene expression, splicing, histone modifications, TF binding, and 3D contact maps from DNA sequence.

What it does

• Inputs: 1-Mb genomic interval + reference/alternate alleles for variant scoring
• Outputs: ATAC, CAGE, DNASE, RNA_SEQ, CHIP_HISTONE, CHIP_TF, SPLICE_SITES, SPLICE_SITE_USAGE, SPLICE_JUNCTIONS, CONTACT_MAPS, PROCAP
• Unique capability for STRC: tissue-resolved splicing-effect quantiles for non-coding/synonymous-region variants where AlphaMissense (protein-level) is silent

How to use

from alphagenome.data import genome
from alphagenome.models import dna_client, variant_scorers
 
dna_model = dna_client.create(api_key="YOUR_KEY")  # Google AI Studio key
 
variant = genome.Variant(
    chromosome="chr15",
    position=43600551,
    reference_bases="T",
    alternate_bases="G",
)
 
interval = variant.reference_interval.resize(dna_client.SEQUENCE_LENGTH_1MB)
scores = dna_model.score_variant(
    interval=interval,
    variant=variant,
    variant_scorers=list(variant_scorers.RECOMMENDED_VARIANT_SCORERS.values()),
)
df = variant_scorers.tidy_scores(scores)

• Install: pip install alphagenome
• Auth: API key from Google AI Studio (same as Gemini API keys), env GOOGLE_API_KEY

STRC usage

• h01 — applied 2026-04-08 to Misha’s E1659A variant (chr15:g.43600551T>G). 54,276 total scores across tissues; 1,528 STRC-specific. Splice junction quantile 0.997+ across ALL tissues (highly significant). Top splice effect: frontal cortex raw_score=0.377, quantile=0.998. Cited in STRC E1659A Computational Tool Audit.

Last result

2026-04-08 — Splice quantile 0.998 (frontal cortex) on E1659A; result CSV at ~/Documents/STRC-AlphaGenome-E1659A.csv.

Connections

• [part-of] Tools index
• [applies] h01 hub
• [see-also] AlphaMissense
• [see-also] STRC E1659A Computational Tool Audit
• [applies] STRC Variant c.4976A>C — Misha