~40M people have DNA chip data sitting unused. Most consumer reports give you 30-300 health traits. We score 1,261+ polygenic risk scores across every major disease category, scan 400,000+ variants against ClinVar for pathogenic findings,run pharmacogenomics on 34 genes (110K+ diplotypes via CPIC), and generate an AI-powered longevity protocol.
How the scoring works:
- Effect allele dosages x published PGS Catalog weights
- Percentiled against empirical distributions from 2,504 1000 Genomes Phase 3 samples across 5 ancestry groups (EUR, AFR,EAS, SAS, AMR)
- Ancestry-aware ML correction via Ridge regression (8 features including ancestry proportions)
- Optional deep imputation (Beagle 5.5) expands 609K chip variants to ~28M
- Each score shows confidence level, variant coverage, and population calibration
What you get:
- 1,261+ polygenic risk scores organized by category (cardiovascular, neurological, metabolic, cancer, autoimmune, etc.)
- ClinVar pathogenic variant matches (400K+ scanned)
- Pharmacogenomics: star allele calling for 34 genes with CPIC dosing recommendations
- Convergence detection: flags when multiple PRS + monogenic variants point at the same disease pathway
- AI-generated longevity protocol (supplements, training, diet — all matched to your genetic profile)
- Interactive genome explorer + 4 downloadable PDFs (Health Summary, Doctor's Report, Full Analysis, Health Protocol)
- JSON export to feed your genetic profile to AI assistants
Privacy (this was the main design constraint):
Zero data retention. No accounts, no cookies, no tracking. Your DNA file is deleted from disk immediately after analysis (~2 minutes). All remaining data (reports, scores, PDFs) is purged within 2 hours. At deletion time, you receive an automated Data Deletion Certificate — a PDF with a SHA-256 attestation hash (SHA256(certificateId:jobId:deletedAt:salt)) proving when every piece of your data was destroyed. Every deletion is logged in an append-only audit trail containing zero genetic data.
We built this specifically because 23andMe's bankruptcy exposed 15 million people's DNA. There is literally nothing to breach here.
Honest caveats:
- PRS capture common variant risk only — they're not crystal balls, they show relative genetic risk vs. the population
- Most PGS models were developed in European cohorts; cross-ancestry performance varies
- This is a research/educational tool, not a certified diagnostic
- Still some rough edges — we've only tested with a handful of people so far
We've had some surprisingly accurate results — one report independently flagged genetic variants consistent with a known medical condition that hadn't been mentioned to us.
Giving away 5 free beta codes to the first commenters. Would genuinely love honest feedback, bug reports, and criticism.