Mmcp.market

alphagenome skill

by K-Dense-AI·K-Dense-AI/scientific-agent-skills·47k stars·MIT

Look up precomputed AlphaGenome Atlas effects for any GRCh38 single-nucleotide variant (AVI score with Phred and 18 SHAP feature attributions, plus raw and quantile scores for RNA-seq, DNase, ATAC, ChIP-TF, ChIP-histone, CAGE, PRO-cap, splicing, polyadenylation and contact-map tracks), score variants or scan windows on demand with the AlphaGenome model for human and mouse (variant scoring, in silico mutagenesis, REF-versus-ALT track prediction), and build Atlas website deep links. Use when the user mentions AlphaGenome, AlphaGenome Atlas, AVI or AlphaGenome Variant Impact, DeepMind variant effect prediction, or wants to prioritise or mechanistically interpret non-coding, regulatory, splicing, enhancer, promoter, or chromatin-accessibility effects of SNVs from a VCF, credible set, or region. Research use only; not a clinical tool.

A100/100content scan

Is the alphagenome skill safe?

Clean: nothing in its files matched our rules. We read 8 files in the folder on 2026-09-28.

No findings.

Install the alphagenome skill

A skill is a folder. Copy it into your agent's skills folder and the agent loads it when the task matches its description. 

git clone --depth 1 https://github.com/K-Dense-AI/scientific-agent-skills.git /tmp/scientific-agent-skills
mkdir -p ~/.claude/skills
cp -r /tmp/scientific-agent-skills/skills/alphagenome ~/.claude/skills/alphagenome
available in every project

In the Claude apps, zip the folder and upload it from the Skills settings. The folder on GitHub

The instructions your agent would load

SKILL.md as published, without the frontmatter. Read it on GitHub

AlphaGenome and the AlphaGenome Atlas

AlphaGenome is DeepMind's sequence-to-function model: 1 Mb of DNA in, base-pair predictions for eleven assay types across thousands of human and mouse tracks out. The AlphaGenome Atlas (released 2026-09-08) is that model run once over every possible single-nucleotide change in GRCh38, about 9 billion variants, stored with a single ranking number, the AlphaGenome Variant Impact (AVI) score, its genome-wide percentile, and an 18-way attribution of what drives it. Both are reached through one pip install alphagenome and one API key.

Research and theoretical modelling only. Outputs must not be used to train

other models, and are not for diagnostic procedures or medical decisions.

When to use which

Setup

uv pip install alphagenome                     # PyPI; tested on Python 3.12 and 3.13, alphagenome 0.9.0
export ALPHAGENOME_API_KEY="..."               # https://deepmind.google.com/science/alphagenome
cd skills/alphagenome/scripts
python atlas_query.py scorers                  # proves key + network in one call

Never put the key on a command line or in a file you commit; the scripts only read it from the environment. An invalid key surfaces as ValueError: API key not valid, not as a permission error.

The coordinate contract

(22-36201698-A-C), GTEx (chr2236201698ACb38), and Open Targets spellings are accepted by the scripts and by genome.Variant.from_str.

  • A variant is 1-based chr:pos:ref>alt (chr22:36201698:A>C). gnomAD

SDK's genome.Interval is 0-based half-open. The scripts convert.

  • An interval on the command line is 1-based closed chr:start-end; the

the reference, so a variant with REF and ALT swapped, or on GRCh37, returns a wrong record silently. Check REF against the FASTA before trusting a lookup.

  • Human is GRCh38 only. The Atlas key is chr:pos:alt; REF is implied by
  • rsIDs are not accepted by the API or the portal. Resolve them to coordinates.
  • Use the chr prefix; MT becomes chrM.

Atlas workflow

1. Rank with AVI

python atlas_query.py avi --variant chr22:36201698:A>C chr9:128225994:G>A
python atlas_query.py avi --input candidates.vcf --min-phred 20 -o avi.tsv
python atlas_query.py avi --interval chr11:5225727-5226575 --top-k 25 -o hbb_window.tsv
python atlas_query.py avi --input credible_set.tsv --with-tracks -o avi_tracks.tsv

Output, one row per variant:

The Atlas report's advice: rank, do not threshold, and pick thresholds by region or application. Pathogenic regulatory variants sit in lower AVI bins than protein-truncating or splice-motif variants, so a single genome-wide cut-off under-calls exactly the variants this resource was built for.

Read the attribution before the number. MERGEDSPLICING or ALPHAMISSENSE on top means a splice or coding mechanism; MAXABSDNASE, MAXABSCHIPTF, MAXABSRNASEQ mean a regulatory mechanism you can resolve by track; CACTUS241WAY or PHASTCONS470_WAY on top means conservation is carrying the score and the molecular mechanism is not resolved.

2. Resolve the mechanism by track

python atlas_query.py scorers                                   # what the server serves right now
python atlas_query.py tracks --scorer RNA_SEQ --query colon     # find ontology CURIEs
python atlas_query.py scores --variant chr22:36201698:A>C \
    --scorers RNA_SEQ DNASE SPLICE_SITE_USAGE --ontology UBERON:0001157 -o colon.tsv
python atlas_query.py scores --interval chr11:5225727-5226575 --scorers CHIP_TF --gene HBB -o hbb_tf.tsv

One row per variant x track (x gene for RNASEQ, POLYADENYLATION, SPLICE), with rawscore and, where served, quantilescore. Track-level scorer names: ATAC, DNASE, CHIPTF, CHIPHISTONE, CAGE, PROCAP, RNASEQ, POLYADENYLATION, SPLICESITES, SPLICESITEUSAGE, SPLICEJUNCTIONS, CONTACTMAPS, plus *_ACTIVE variants; scorers is the authority on the live list. Filter by the tissue the question is about, not by the genome-wide maximum: 9,440 tracks means something is always extreme somewhere.

3. Send the reader to the portal

python atlas_link.py variant chr22:36201698:A>C --biosample "colon" --modalities RNA_SEQ,DNASE,CHIP_TF
python atlas_link.py locus chr11:5225727-5226575 --tf GATA1
python atlas_link.py gene HBB --markdown

No key, no network. The site shows the AVI track, per-modality heatmaps over every biosample, REF-vs-ALT prediction tracks, and motif instances. Attach a link to every variant you report.

In Python

import os
from alphagenome.atlas import atlas
from alphagenome.data import genome

client = atlas.create(os.environ["ALPHAGENOME_API_KEY"], timeout=30)
scores = client.query_variant(
    genome.Variant.from_str("chr22:36201698:A>C"),
    requested_scorers=["AVI_SCORE", "AVI_SCORE_FEATURE_IMPORTANCE", "RNA_SEQ"],
    ontology_terms=["UBERON:0001157"],          # optional; ignored for scorers without ontology metadata
)
avi = scores["AVI_SCORE"]                        # AnnData: X (1,1) raw; layers['quantiles'] (1,1) cdf
fi = scores["AVI_SCORE_FEATURE_IMPORTANCE"]      # AnnData: X (1,18); var['name'] = feature keys
rna = scores["RNA_SEQ"]                          # AnnData: obs = variant x gene, var = tracks, X = log2 FC
client.query_interval(genome.Interval("chr11", 5225726, 5226575), requested_scorers=["AVI_SCORE"])

queryinterval returns all 3 SNVs per base, in 32 bp chunks. Keep windows to about 1 kb (3,000 variants); atlasquery.py refuses more unless --max-window is raised. query_variants stops at the first failed lookup; the script queries one variant at a time so misses become error cells.

Model workflow

Score variants the Atlas does not hold

python score_variants.py --variant chr22:36201698:A>C -o scores.tsv                   # 12 recommended scorers, 1 Mb
python score_variants.py --input indels.vcf --scorers RNA_SEQ SPLICE_SITE_USAGE \
    --ontology UBERON:0001157 --min-abs-quantile 0.99 -o colon.tsv
python score_variants.py --organism mouse --variant chr7:45000000:A>G --sequence-length 500KB
python score_variants.py --list-scorers
python score_variants.py --list-tracks --output-type RNA_SEQ --query liver -o tracks.tsv

Output is the official tidy table from variantscorers.tidyscores: one row per variant x scorer x track (x gene) with rawscore and quantilescore, sorted by |raw|. Default scorers are the 12 recommended difference scorers; --include-active adds the seven *_ACTIVE activity scorers. At most 20 scorers per request.

from alphagenome.models import dna_client, variant_scorers
model = dna_client.create(os.environ["ALPHAGENOME_API_KEY"])
variant = genome.Variant.from_str("chr22:36201698:A>C")
interval = variant.reference_interval.resize(dna_client.SEQUENCE_LENGTH_1MB)
adatas = model.score_variant(interval, variant, variant_scorers=[variant_scorers.RECOMMENDED_VARIANT_SCORERS["RNA_SEQ"]])
df = variant_scorers.tidy_scores(adatas)         # filter df.ontology_curie afterwards; score_variant takes no ontology_terms

Predict tracks and mutagenise

vo = model.predict_variant(interval, variant,
                           requested_outputs=[dna_client.OutputType.RNA_SEQ, dna_client.OutputType.DNASE],
                           ontology_terms=["UBERON:0001157"])
vo.reference.rna_seq.values, vo.alternate.rna_seq.values      # (1048576, n_tracks)

window = genome.Interval("chr20", 3_753_000, 3_753_400).resize(dna_client.SEQUENCE_LENGTH_16KB)
ism = model.score_ism_variants(interval=window, ism_interval=window.resize(256),
                               variant_scorers=[variant_scorers.CenterMaskScorer(
                                   requested_output=dna_client.OutputType.DNASE, width=501,
                                   aggregation_type=variant_scorers.AggregationType.DIFF_MEAN)])

Supported windows: 16 kb, 100 kb, 500 kb, 1 Mb (214 to 220); 1 Mb is the default and is required for distal enhancers and contact maps. Ontology terms are CURIEs (UBERON:0002048 lung, CL:0000084 T cell); discover them with --list-tracks or model.output_metadata(...).concatenate(). Plotting, gene annotation (GENCODE v46 Feather on GCS), splicing and haplotype recipes: references/model-api.md.

Reading the numbers

Always report raw score and quantile or Phred, with the scorer, track, biosample CURIE, and gene. rawscore is the effect size on the scorer's scale (RNASEQ is log2 fold change: -1 is half); quantilescore is the rank against common variants and saturates near 0.99999. A quantile above 0.99 with |raw| < 0.1 is the standard artefact of a quiet region and means no effect. Unsigned scorers (SPLICE, POLYADENYLATION, CONTACTMAPS, ACTIVE) have no direction. Most variants are benign; "AlphaGenome predicts no molecular effect" is a complete answer, and a variant inside a peak whose REF and ALT tracks are identical is not "disrupting" anything. Full rules, tissue matching, and the reporting checklist: references/interpretation.md.

What the model cannot see: trans effects, non-polyadenylated RNAs (snRNA genes such as RNU4-2), cell types absent from training, protein-level consequences (AlphaMissense is folded into AVI for that), RNA structure and miRNA biology, diploid dosage, developmental time, species other than human and mouse.

Limits, quota, terms

promised later. Reference N bases were never scored.

  • Atlas: GRCh38 SNVs only for now; indels were scored for the paper and are

larger query rate than on-demand prediction. Transient RESOURCE_EXHAUSTED and UNAVAILABLE are retried by the client (5 attempts, back-off to 60 s).

  • Quotas are per key and unpublished; the Atlas is documented as having a

Tabix download at https://alphagenome.google/downloads; feature attributions and splicing scores are non-commercial downloads; all other raw track scores are API-only and non-commercial. Commercial API access is "coming soon" via Google Cloud Model Garden.

  • Access tiers (Atlas report): AVI scores are also a permissively licensed

DeepMind's terms. Cite Avsec et al., Nature 649:1206 (2026) and the Atlas report (Cheng, Taylor, Nicolaisen, Pan, Bycroft, Perino, Ward et al., 2026).

  • The alphagenome client is Apache-2.0; model weights and outputs carry

References

configurations with track counts, AVI training and the 18 features, quantile to Phred, the client API and AnnData layout, error mapping, access tiers, portal URL grammar, GTF and download locations.

  • references/atlas.md - what the Atlas contains, the 19 scorer

lengths, output types and track counts, ontology metadata, predict and score calls, recommended scorer configurations, ISM, gene annotation, plotting.

  • references/model-api.md - dna_client cheat sheet: coordinates, sequence

tissue matching, negative results, model blind spots, coordinate hygiene, reporting checklist.

  • references/interpretation.md - raw versus quantile, AVI thresholds,

tracks), scripts/scorevariants.py (model scoring, --list-scorers, --list-tracks), scripts/atlaslink.py (portal deep links, offline).

  • Scripts: scripts/atlas_query.py (Atlas: avi, scores, scorers,

More skills from K-Dense-AI/scientific-agent-skills

  • AadaptyvHow to use the Adaptyv Bio Foundry API and Python SDK for protein experiment design, submission, and results retrieval. Use this skill whenever the user mentions Adaptyv, Foundry API, protein binding assays, protein screening experiments, BLI/SPR assays, thermostability assays, or wants to submit protein sequences for experimental characterization. Also trigger when code imports `adaptyv`, `adaptyv_sdk`, or `FoundryClient`, or references `foundry-api-public.adaptyvbio.com`.
  • AaeonThis skill should be used for time series machine learning tasks including classification, regression, clustering, forecasting, anomaly detection, segmentation, and similarity search. Use when working with temporal data, sequential patterns, or time-indexed observations requiring specialized algorithms beyond standard ML approaches. Particularly suited for univariate and multivariate time series analysis with scikit-learn compatible APIs.
  • Aanalytical-method-validationPlan, execute, and document validation, verification, and transfer of analytical procedures under the governing framework - ICH Q2(R2) and Q14, USP <1220>/<1225>/<1226>, ICH M10 bioanalytical, CLSI EP, or ISO/IEC 17025. Use for HPLC, LC-MS/MS, GC, CE, ICP-MS, dissolution, qNMR, qPCR, NIR, and ligand binding or cell-based assays whenever the question is whether a procedure is fit for its intended purpose. Triggers include "method validation", "analytical method validation", "AMV", "validation protocol", "acceptance criteria", "linearity", "reportable range", "accuracy and precision", "repeatability", "intermediate precision", "recovery", "LOD", "LOQ", "detection limit", "quantitation limit", "specificity", "robustness", "method transfer", "method comparison", "Deming", "Passing-Bablok", "Bland-Altman", "equivalence testing", "OOS investigation", "ICH Q2", "Q2(R2)", "Q14", "USP 1225", "ICH M10", "incurred sample reanalysis", "ISR", "CLSI EP", and any request to show that an assay works.
  • AanndataData structure for annotated matrices in single-cell analysis. Use when working with .h5ad files or integrating with the scverse ecosystem. This is the data format skill—for analysis workflows use scanpy; for probabilistic models use scvi-tools; for population-scale queries use cellxgene-census.
  • AarborAutonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. "get my model's eval score up", "improve this agent/harness", "tune this pipeline", "beat the baseline on this benchmark", "run a search over approaches and keep the best", "do an MLE-bench / Kaggle-style optimization", or any long-horizon "make this artifact better and don't just memorize the dev set" task. Trigger it even when the user doesn't say "Arbor" or "hypothesis tree" but describes repeated experiment-and-evaluate loops, branching exploration of competing ideas, or worries about a dev/test gap. Runs Claude itself as the coordinator with subagent executors in isolated git worktrees; for the standalone `arbor` CLI tool see references/arbor-upstream.md.
  • AarboretoInfer gene regulatory networks (GRNs) from gene expression data using scalable algorithms (GRNBoost2, GENIE3). Use when analyzing transcriptomics data (bulk RNA-seq, single-cell RNA-seq) to identify transcription factor-target gene relationships and regulatory interactions. Supports distributed computation for large-scale datasets.
  • AastropyCore Python library for astronomy and astrophysics workflows that need Astropy APIs, including units/quantities, coordinates, FITS I/O, tables, time systems, WCS, and cosmology. Use when implementing or debugging astronomical data analysis code with Astropy.
  • AautoskillObserve the user's screen via screenpipe, detect repeated research workflows, match them against existing scientific-agent-skills, and draft new skills (or composition recipes that chain existing ones) for the patterns not yet covered. Use when the user asks to analyze their recent work and propose skills based on what they actually do. Requires the screenpipe daemon (https://github.com/screenpipe/screenpipe) running locally on port 3030 — the skill has no other data source and will refuse to run if screenpipe is unreachable. All detection runs locally; only redacted cluster summaries reach the LLM.
  • Abenchling-integrationBenchling Python SDK and REST API integration for registry entities, inventory, ELN entries, workflows, Benchling Apps, and Data Warehouse queries. Use when automating lab data with benchling-sdk or the v2 API.
  • Abgpt-paper-searchSearch scientific papers and retrieve structured experimental data extracted from full-text studies via the BGPT MCP server. Returns 25+ fields per paper including methods, results, sample sizes, quality scores, and conclusions. Use for literature reviews, evidence synthesis, and finding experimental details not available in abstracts alone.
  • AbidsUse this skill when working with Brain Imaging Data Structure (BIDS) datasets: organizing neuroscience and biomedical data (MRI, EEG, MEG, iEEG, PET, microscopy, NIRS, motion capture, EMG, MR spectroscopy, behavioral), querying BIDS layouts, validating compliance, converting DICOM to BIDS, writing metadata sidecars, or creating BIDS derivatives.
  • AbiopythonComprehensive molecular biology toolkit. Use for sequence manipulation, file parsing (FASTA/GenBank/PDB), phylogenetics, and programmatic NCBI/PubMed access (Bio.Entrez). Best for batch processing, custom bioinformatics pipelines, BLAST automation. For quick lookups use gget; for multi-service integration use bioservices.

All agent skills → · MCP servers