skills.rest
K-Dense-AI
K-Dense-AIOFFICIAL
Explore all skills by this author →
Website𝕏@k_dense_aiGitHub

alphagenome

Query precomputed AlphaGenome Atlas variant impact scores and predict regulatory effects of GRCh38 SNVs.

46.6k|4.2k|Updated Oct 19, 2025
Search Tags:#genomics#vcf#snv#variant effect prediction#alphagenome#deepmind#regulatory variants
One-click install
npx skills add https://github.com/K-Dense-AI/claude-scientific-skills --skill alphagenome-k-dense-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: alphagenome
Source: https://github.com/K-Dense-AI/claude-scientific-skills/tree/main/skills/alphagenome
Command: npx skills add https://github.com/K-Dense-AI/claude-scientific-skills --skill alphagenome-k-dense-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires alphagenome, numpy, pandas, anndata, grpcio, and includes scripts (resource) and references (resource) components.

What problem does it solve? Prioritizing and mechanistically interpreting non-coding, regulatory, and splicing variants from VCFs or credible sets requires running large sequence-to-function models; this Skill provides direct access to DeepMind's AlphaGenome Atlas (precomputed effects for ~9 billion GRCh38 SNVs) and the on-demand AlphaGenome model without writing client code. ## Core Features & Use Cases - Atlas AVI Ranking: Look up the AlphaGenome Variant Impact (AVI) score, Phred-scaled genome-wide percentile, and 18 SHAP feature attributions for any GRCh38 SNV, VCF, or window up to ~1 kb. - Track-Level Mechanism Resolution: Retrieve raw and quantile scores across RNA-seq, DNase, ATAC, ChIP-TF, ChIP-histone, CAGE, PRO-cap, splicing, polyadenylation, and contact-map tracks filtered by tissue ontology CURIE or gene. - On-Demand Model Scoring: Score indels, mouse (mm10) variants, custom windows, or run in silico mutagenesis and REF-versus-ALT track prediction with the AlphaGenome model. - Portal Deep Links: Build AlphaGenome Atlas website URLs for any variant, locus, or gene so readers can inspect prediction tracks. - Use Case: Given a GWAS credible set in a VCF, rank every variant by AVI Phred, identify whether splicing or TF-binding drives the top hit, pull colon RNA-seq scores for the lead variant, and attach a portal link to the report. ## Quick Start Ask the AI to look up the AlphaGenome Atlas AVI score and feature attributions for the variant chr22:36201698:A>C and explain which molecular mechanism drives it.

Frequently Asked Questions about alphagenome

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get the AlphaGenome AVI score for a variant?▼

Run scripts/atlas_query.py with the avi subcommand and a variant in chr:pos:ref>alt form, for example chr22:36201698:A>C. The output includes the raw AVI score, Phred-scaled percentile, and all 18 SHAP feature attributions, and requires an ALPHAGENOME_API_KEY environment variable.

What is the difference between the AlphaGenome Atlas and the AlphaGenome model?▼

The Atlas holds precomputed scores for every possible GRCh38 single-nucleotide variant with higher query quotas, while the model scores anything the Atlas lacks: indels, mouse mm10 variants, custom windows, and REF-versus-ALT track predictions. Both use the same alphagenome Python package and API key.

Does AlphaGenome support GRCh37 coordinates or rsIDs?▼

No. The Atlas is GRCh38-only and neither the API nor the portal accepts rsIDs. Lift GRCh37 coordinates over and resolve rsIDs to coordinates first, and verify the REF allele against the reference FASTA because a swapped REF/ALT silently returns a wrong record.

Can AlphaGenome score indels and mouse variants?▼

The public Atlas API currently serves SNVs only, but the on-demand model scores indels and mouse mm10 variants through scripts/score_variants.py with --organism mouse or an indel VCF. Indel support in the Atlas is promised for a later release.

Why does a variant show a high quantile score but a tiny raw score?▼

In low-expression genes and quiet regions the background distribution is so narrow that a raw change below 0.1 can rank above the 0.99 quantile. This is a standard artefact meaning no real effect; always report raw score and quantile together.

Is AlphaGenome suitable for clinical variant interpretation?▼

No. AlphaGenome outputs are research-use-only predictions of molecular effects, explicitly excluded from diagnostic procedures and medical decisions by the terms of use. A high AVI score is one evidence line, never a pathogenicity verdict.

Related Skills

Explore More Education & Research Skills→
openaiopenai
5.3k

gtex-eqtl-skill

Fetch GTEx single-tissue eQTL associations for one variant as JSON.

Official
Intermediate
CentauriounCentaurioun

tooluniverse-rare-disease-diagnosis

Map phenotypes to HPO terms and rank rare disease diagnoses.

Community
Advanced
exon-researchexon-research
475

genomi-gnomad

Fetch population allele frequencies for a specified variant from gnomAD.

Official
Basic
nahisahonahisaho
3

scientific-variant-effect-prediction

Integrate AlphaMissense, CADD, and SpliceAI to score and classify variant pathogenicity.

Community
Advanced