vartriage

A streaming pipeline for identifying and classifying pathogenic genetic variants from VCF data. Processes whole-genome scale files (4M+ variants) under 2GB memory via batched iterators.

Reads a VCF, applies quality filters, annotates functional consequence and population frequency, computes pathogenicity scores, runs ACMG/AMP evidence classification, and writes a ranked candidate list in JSON, CSV, or PDF.

Install

pip install vartriage

Quick start

from pathlib import Path
from vartriage import Pipeline, PipelineConfig, AnnotationConfig

config = PipelineConfig(
    vcf_path=Path("sample.vcf.gz"),
    output_path=Path("candidates.json"),
    annotation=AnnotationConfig(
        gene_annotation_path=Path("gencode.v44.gtf"),
        gnomad_path=Path("gnomad.v4.sites.tsv"),
    ),
)

pipeline = Pipeline(config)
pipeline.run()

See Getting Started for installation options and a full walkthrough.

For zero-config annotation without local reference files, see API Mode.

For multi-sample cohort analysis (shared variants, gene burden), see Cohort Analysis.

For mitochondrial DNA variant analysis (heteroplasmy, MITOMAP, mtDNA-specific classification), see Mitochondrial Variants.

For pre-flight sample quality control (Ti/Tv, het/hom, variant count sanity checks with a strict gate), see Quality Control.

For the full ACMG/AMP classification reference (criteria, thresholds, PS1/PM5 setup, combining rules), see ACMG Criteria.