A local-first genomic analysis toolkit

I built this over the weekend after downloading my raw file and realising how much clinically significant data was in there that the company never surfaced for me.

The tool parses raw genotype files from 23andMe or AncestryDNA and does the following:

Cross-references ~1,500 curated, clinically significant SNP variants across 21 categories (cardiovascular, metabolic, autoimmune, pharmacogenomic, etc.) Detects convergence across 22 biological pathways — individual SNPs are noisy, but when you have multiple variants across lipid metabolism, coagulation, and cardiovascular pathways all feeding into the same risk, that's a signal. The tool uses sigmoid synergy scoring to surface those patterns. Calls pharmacogenomic metabolizer phenotypes for 11 genes (CYP2D6, CYP2C19, CYP2C9, CYP3A4, CYP3A5, CYP1A2, DPYD, TPMT, SLCO1B1, UGT1A1, ABCB1) with CPIC-aligned star-allele calling, and maps 60+ drug interactions with specific dosing actions. Computes polygenic risk scores with z-score → percentile conversion. Generates a full narrative report (markdown or JSON), a GP-readable summary card, and actionable recommendations. Includes an interactive React dashboard with a body map, variant table, pathway cards, and risk timeline.

Privacy was the first design decision. Your genome is the most sensitive data you have — it doesn't change, it can't be rotated, and it identifies you and your family permanently. Everything runs locally. No server, no API calls during analysis, no telemetry. Your data stays on your machine.