Benchmark · Real chemistry ·

Catch medication risk before
the prescription reaches the patient.

BioReason Health is building the auditable clinical-flag layer for Indian diagnostic labs and hospitals. Two live wedges: MedSafe for pre-prescription medication risk and PetriDish Autoimmune for lab-report triage.

Deterministic rule engines. No LLM in the disposition path. Every reason cites its source guideline or approved product information. Missing critical data blocks the disposition rather than producing false reassurance.

Try MedSafe →Paste an autoimmune lab reportStart a validation pilot
github.com/shailesh2790/petridish · Apache 2.0
Architecture

Ingest · Interpret · Discover

01
Ingest
VCF · 23andMe · CSV · any lab format
02
Interpret
Deterministic router · India-calibrated tables
03
Discover
Hypothesis-quarantined layers · CRO-ready outputs
Platform

Productized capabilities · one shared trust layer

All labelled with validation state; research modules under /research
PetriDish MedSafe
Guideline implemented
Pre-prescription medication risk stratification
CPIC-encoded rules · neutral 5-state disposition · missing-data first-class
PetriDish Autoimmune · Lab Reader
Guideline implemented
Messy autoimmune lab report → auditable disposition
LLM extracts · deterministic rules decide · every flag cites the source line
PetriDish Autoimmune · Navigator
Guideline implemented
Structured symptom + prior-lab intake · deterministic 4-tier flag
Red / Amber / Yellow / Abstain · rule-versioned · clinician-review mandatory
PetriDish Rx
Guideline implemented
Pharmacogenomic decision support
CPIC-graded variants · India-calibrated tables
PetriDish Lens
Guideline implemented
ANA-IIF pattern → reflex genotype
29 ICAP AC-codes · wired to Rx + Immune
PetriDish Immune
Guideline implemented
HLA precision medicine
Hard benchmark gate · SJS/TEN discipline
Who this is for
For Pharma R&D
Population-specific epitope + candidate prioritization for antigen-specific Treg and tolerizing-vaccine programs. Real Rammensee anchor-motif scoring. IEDB-validated substrate. India-calibrated HLA distributions no Western tool captures.
TregEpitope Intelligence
For Diagnostic Labs
Your PGx panel returns star alleles. We turn them into a clinician-actionable, source-cited, co-branded report. You keep the lab relationship and branding; we provide the interpretation, audit trail and updateable rule engine.
Co-Branded Report
For Hospitals & Pharmacies
Pre-prescription medication risk stratification for selected NSAIDs, codeine, tramadol and paracetamol. Neutral 5-state disposition, evidence-cited, clinician-review mandatory.
PetriDish MedSafe
For Academic Research
Grant-research pipeline with live Ensembl VEP + gnomAD + ClinVar APIs and IndiGen frequency layer. Every score sourced. BIRAC / DBT / Wellcome-ready.
IndoVarClassifier
Built for enterprise

Trust as architecture, not a checkbox

Deterministic core
No LLM in the clinical routing path. Same input, same output, every call.
Audit by design
Every response carries a clickable trace_id. Reasoning chain retrievable in one call.
Predictions never masquerade
Hypothesis layers are structurally quarantined from facts. Two edge types. Zero code paths where they meet.
Open source · Apache 2.0
github.com/shailesh2790/petridish — every method cited, every backend swap-in named.
The India edge

Sourced facts only

Every stat has a citation, a defined population, and a date
25.76%
HLA-DRB1 Shared Epitope carriage
n=67,888 Indian donors · DKMS registry · this repo, 2026
src: scripts/se_carriage_paper_draft.md
*10:01
Dominates Indian SE (37–62% of SE haplotypes)
vs *04:01-dominated Caucasian composition · same cohort
src: scripts/se_carriage_paper_draft.md
NUDT15
Primary Indian thiopurine-safety gene
Elevated LOF prevalence in South Asians vs Europeans
src: CPIC · TPMT/NUDT15 & thiopurines
CYP2C9 *3
Encoded in MedSafe · ibuprofen and celecoxib rules
CPIC 2020a · verified against latest guideline before deploy
src: CPIC · CYP2C9 & NSAIDs (PMID 32189324)

Software calibrated to Indian genetics.

Live at bioreason-india.vercel.app. Open source at github.com/shailesh2790/petridish.

Request AccessLive case studyGitHub