We've released Just-DNA-Lite, an open-source (AGPL-3.0) tool for annotating personal genomes, a from-scratch rewrite of our earlier Just-DNA-Seq.
It takes a GRCh38 VCF/VCF.gz (WGS or WES) and produces variant annotations, polygenic risk scores, and health-risk reports. The pipeline runs on Dagster with Polars/DuckDB engines and VCF reading via polars-bio. A genome that took hours under the old OakVar-based version now annotates in minutes on a laptop. Everything runs locally, no data leaves the machine. Output is Parquet, with annotated VCF export also available.
A few things that may be of interest:
- 5000+ PRS from the PGS Catalog, percentile-ranked against the five 1000 Genomes superpopulations. Scoring runs through our just-prs library (DuckDB or Polars backends), validated against PLINK2 at r = 0.9999.
- AI-assisted module creator that turns a paper or a PGS Catalog score into a deployable annotation module (a YAML spec plus variants/studies tables compiled to Parquet). Useful for prototyping a filter set quickly; outputs still need expert review.
- Pure-Python Reflex web UI, plus CLI and Docker/Podman options.
Current limitations: GRCh38 only (GRCh37/hg19 and microarray inputs like 23andMe are not yet supported), and it is research-use-only, not a clinical or diagnostic tool.
Code: https://github.com/dna-seq/just-dna-lite Demo: https://lite.just-dna.life/ (loaded with two public test genomes you can explore)
Two short video walkthroughs:
- General intro by Livia Zaharia, run on her own genome:
- Agentic annotation-module creation by Newton Winter:
Feedback, bug reports, and suggestions are very welcome, especially on the PRS methodology and the annotation module format.
0 answers
No answers yet.
Log in to answer this question.