IndoPGxlive beta · Groq + Neo4jCPIC-grounded

Standard drug dosing is wrong for 300 million Indians.

IndoPGx is a pharmacogenomics dosing-safety layer calibrated to Indian variant frequencies. Genotype in, CPIC-graded dosing guidance out — in under 200 milliseconds, every recommendation auditable to its source.

The problem nobody is dosing for

Guidelines are built on Western cohorts. The frequencies that decide a dose are different here — and no one is pricing that difference into the prescription.

23%
South Asians carry CYP2C19 loss-of-function vs 15% globally — clopidogrel under-protects them
~3×
Higher thiopurine toxicity risk in NUDT15-variant patients, common in Indian populations
8–15%
G6PD deficiency in parts of India — a known but routinely unchecked prescribing hazard
~0
Global dosing tools that use Indian allele frequencies as their baseline

What a dosing check returns

Every output names its triggering variant, its CPIC evidence grade, and its Indian frequency — and is written to an immutable audit log. The reasoning model may never name a drug, gene, or variant that isn't in the underlying graph result.

POST/pgx/check·CYP2C19 *2/*2·clopidogrel187 ms
Avoid standard dose — Poor metaboliser, reduced clopidogrel activation
This patient cannot convert clopidogrel to its active form efficiently. Expect inadequate platelet inhibition and elevated cardiovascular risk if standard dosing is used.
Consider alternative antiplatelet (prasugrel or ticagrelor) where not contraindicated, per CPIC guidance for CYP2C19 poor metabolisers.
CPIC guideline · level 1A
India freq: 23% (IndiGen)
logged to audit trail

How it works

1
Send genotype
Diplotype or star-allele for the gene of interest, plus the drug. One API call or a CSV batch.
2
Engine resolves
CPIC rule engine maps genotype to phenotype; the graph overlays the Indian frequency context.
3
Graded guidance
A dosing recommendation with evidence grade, rationale, and audit record returns in under 200 ms.

Who it's for

🧪
Diagnostic labs
Add an India-calibrated interpretation layer to PGx panels. Turn a raw genotype report into a clinician-ready dosing report.
Per-sample API
🏥
Hospitals & CMIOs
Embed /pgx/check in the prescribing flow. Surface a dosing alert at order entry. Measure alert acceptance and avoided adverse events.
EHR embed
💊
Pharma & CROs
Stratify Indian trial cohorts by metaboliser status. De-risk dosing arms with population-accurate frequencies before enrolment.
Trial design

Why it's trustworthy

CPIC-grounded
Recommendations trace to published CPIC guidelines with their evidence level shown.
IndiGen frequencies
Variant context drawn from CSIR-IGIB Indian population data, not Western proxies.
Auditable
Every protected action is logged server-side with a timestamp for review.
No hallucinated entities
The model is contractually bound to the graph result rows. No row, no claim.

Use it

Three ways to deploy IndoPGx, from free interactive checks to embedded EHR alerts.

OPEN BETA
Free interactive checks
Open the demo. Run a check. See the trust triad live. No login.
LAB API · revenue wedge
Per-sample interpretation
For diagnostic labs running PGx panels. Add the India-calibrated interpretation layer. You generate the genotype — we turn it into a clinician-ready dosing report.
HOSPITAL EMBED
EHR prescribing alerts
Annual licence to embed dosing checks in your prescribing flow. Seat or volume-based.

Ready to ship safer dosing in India?

A pilot starts with a single dosing-check on your data. From there: a calibration review, a per-sample contract, or an EHR-embed scoping call.

API reference
Endpoints · request/response schema · curl + Python + JS examples · rate limits.
Dependency licensing
Every shipped dependency · commercial-use status verified · non-commercial firewalled.
Indian data layer
The moat · sources · IndiGen commercial licensing status disclosed openly.
Engine status
Live API version · SLA targets · audit-log policy · grounding-contract enforcement.
Live data state
Graph counts straight from /api/stats — gaps disclosed.
Decision-support positioning, not a medical device claim. v1.1.0 · CPIC-2024 corpus.