Skip to content

AlphaGenome Workflow

The AlphaGenome workflow wraps sequence analysis, variant-effect analysis, integration helpers, and output track discovery.

CLI

genomics alphagenome analyze -- -i sequence.fasta -k API_KEY -o results/
genomics alphagenome integrate -- --integrated --vcf vcf/sample.vcf.gz --ref refs/GRCh38.fa --api-key API_KEY --output integrated_analysis/
genomics alphagenome tracks --api-key API_KEY --output configs/workflows/alphagenome/tracks.json

Arguments after -- are forwarded to the underlying workflow modules.

Every command reaches AlphaGenome through genomics.core.alphagenome_connection.create_dna_client: the hosted API with ALPHAGENOME_API_KEY (or --api-key), or a self-hosted server when ALPHAGENOME_ADDRESS is set (no key needed). To run the model on this machine's GPU:

genomics alphagenome server setup && genomics alphagenome server start
export ALPHAGENOME_ADDRESS=grpc://127.0.0.1:50051

See Requirements and Visualizer → AlphaGenome backend.

Code And Configs

Kind Path
Code src/genomics/workflows/alphagenome/
Configs configs/workflows/alphagenome/
Sequence and output guide docs/guides/alphagenome-sequences.md
Track discovery script scripts/diagnostics/check_alphagenome_outputs.py

Typical Uses

  • predict functional genomic features from FASTA sequences;
  • compare variant effects;
  • export output ontology metadata;
  • integrate VCF/BED/gene-list regions with AlphaGenome predictions.

Implementation Structure

Module Responsibility
neural_module.py Main AlphaGenome sequence/variant analysis workflow
neural_integration.py Extracts regions from VCF/BED/gene lists and invokes analysis
alphagenome_output_tracks.py Queries/export available AlphaGenome tracks and ontology metadata
neural_visualizations_advanced.py Plotting and comparison helpers

Imports of the external alphagenome package are intentionally lazy. This keeps genomics --help and basic package imports working in environments where AlphaGenome is not installed.

Analyze Flow

FASTA input
  -> sequence validation and size checks
  -> AlphaGenome model request
  -> selected output tracks
  -> arrays, metadata, plots, and optional variant comparison

The workflow supports the sequence sizes accepted by AlphaGenome. See AlphaGenome Sequences And Outputs for practical extraction, sizing, and output selection examples.

Integration Flow

The integration command can derive sequences from variant or region inputs and run AlphaGenome analysis in one pass.

Supported sources include:

  • VCF regions;
  • BED intervals;
  • gene lists;
  • manually supplied FASTA files.

Use genomics alphagenome tracks to export the currently available AlphaGenome output and ontology metadata:

genomics alphagenome tracks --api-key API_KEY --output configs/workflows/alphagenome/tracks.json

That metadata is useful when selecting alphagenome_outputs and ontology_terms for dataset builders and genotype predictor configs.

Outputs

Outputs depend on selected tracks and options, but commonly include:

  • NumPy arrays for predicted tracks;
  • CSV/JSON metadata;
  • plots comparing reference and alternate sequence predictions;
  • summaries of available output ontology labels.