Pipelines

The visualizer is the front door. Behind it are command-line pipelines that do the heavy lifting and can be scripted, scheduled or run on a cluster. Most visualizer forms start one of them as a background job, and each job stores the exact command it ran (results/visualizer/jobs/<id>/task.json).

flowchart LR
    subgraph data [Build data]
      GA["genomes-analyzer<br/>FASTQ → VCF"]
      DB["dataset-builders<br/>VCF → canonical dataset"]
      AG["alphagenome<br/>predict-dataset"]
    end
    subgraph learn [Learn from it]
      GT["genotype<br/>CNNs over tracks"]
      VT["variant<br/>transformer"]
      SNP["snp-ancestry"]
    end
    GA --> DB --> AG --> GT
    DB --> VT
    DB --> SNP

Pipeline	CLI	Visualizer equivalent	Code
Dataset builders	genomics dataset-builders	Import dataset	src/genomics/workflows/dataset_builders/
AlphaGenome	genomics alphagenome	AlphaGenome predictions, AlphaGenome page	src/genomics/workflows/alphagenome/
Genotype predictor	genomics genotype	Experiments, Perturbation Lab	src/genomics/predictors/genotype_based/
Variant transformer	genomics variant	–	src/genomics/predictors/variant_transformer/
SNP ancestry	genomics snp-ancestry	–	src/genomics/predictors/snp_ancestry/
Genomes analyzer	genomics genomes-analyzer	–	src/genomics/workflows/genomes_analyzer/
VCF to 23andMe	genomics convert vcf-to-23andme	–	src/genomics/converters/vcf_to_23andme/
Native and third-party	–	–	native/, third_party/
Legacy	–	–	legacy/

To reproduce the 1000 Genomes results from the command line, follow Building 1000 Genomes dataset, then Training and running predictions. Every subcommand is listed in the CLI reference.

All active Python code imports through genomics.* or relative imports inside the same package.