5-minute walkthroughNo login requiredFor investors, hospital CMIOs, biotech BD leads

PetriDish in 5 minutes

Three real Indian clinical wedges PetriDish catches before the standard workflow. Click any card to see the evidence, the metrics, and the live module behind it. Total reading time: ~5 minutes. Live system is one click below each card.

90,650 nodes · 4.4M relationships

55 CPIC-graded PGx variants

100 IMPPAT phytochemicals · 50 curated epifactors

The India wedge

Every existing PetriDish endpoint takes structured JSON inputs (diplotype, HLA allele, rsID list). Doctors don't think in JSON — they think in files. Drop a VCF from Illumina, an Ion Torrent report, or a raw 23andMe export. The Genomic Report Reader auto-detects format, extracts the variants, cross-references against 61 curated CPIC-graded PGx rules, and returns per-gene safety findings sorted by severity. Reused the entire IndoPGx safety engine as the interpretation layer — this module is purely ingestion + normalization.

Supported formats

Actionable PGx variants

Gene families indexed

<60s

Target latency

Reviewer feedback from the June 2026 assessment flagged this exact gap: 'the killer workflow is upload a genomic report → clinician-ready drug safety report in under 60 seconds.' Now shipped. Supports VCF (GT field extraction from sample column), 23andMe (rsid/chr/pos/genotype 4-column format with per-variant ref/alt inference for 15 known PGx rsIDs), and CSV/TSV (rsID required, genotype or carrier_state optional). Actionable findings only — homozygous-reference variants filtered out. 25 MB size cap; larger cohort uploads via /pilot.

Evidence trail

petridishnextversion.pdf

External reviewer's Feature #2: upload 23andMe/GenomeIndia/VCF → explain variants

PGX_RULES

61 CPIC-graded variants with India frequencies from patient.py

VCF spec 4.2

GT field extraction from FORMAT/sample column

23andMe raw

Standard 4-column tab-separated export format

Want to go deeper?

See live data state, knowledge-graph composition, and the validation roadmap.

Truth-in-claims: PetriDish is decision-support software, not a regulatory-approved diagnostic. Clinical recommendations cite their evidence grade and PMIDs. Indian frequencies blend IndiGen 1029-genome, GenomeAsia 100K, and published Indian cohort PMIDs — exact provenance is on /methods. No clinical validation paper yet — that's the next milestone.