LydiaRx
Governed research access · Life-sciences AI

Answers, not data.

Foundation models can predict what a piece of DNA does. They cannot explain why. They cannot turn a population's genetics into a partner-ready drug target. We build the layer that does, for the populations that hold the most drug-relevant biology. Research partners can use it without the data ever moving.

Three offers

What we make available.

01

Governed research access

In-country, purpose-bound, metered access to consanguineous South Asian and Eurasian cohorts for target validation and discovery: rare-variant burden testing, human-knockout catalogues, founder-variant discovery and ancestry-calibrated risk models. Every computation is logged, consent-checked and auditable. Access to a national cohort opens only once that programme's government agreements are in place.

02

Interpretation with provenance

Clinical-grade variant interpretation with explicit ACMG/AMP evidence rather than opaque scores, and decision-grade target-validation dossiers assembled from a provenanced knowledge graph. Every model call and reasoning trace is written to the provenance layer, so what a dossier asserts can always be traced to a computation.

03

Regulatory intelligence

The same evidence engine applied to the regulatory record: thousands of pages of guidance, precedent and submission history refined into decision-grade knowledge that development, regulatory-affairs, medical and commercial teams can each act on and defend.

LydiaRx is currently delivering an AI-strategy and regulatory-intelligence engagement of this kind for a clinical-stage cardiometabolic company.

How access works

Six steps, nothing raw leaves.

1 Apply Purpose, analysis,consent tiers, fee.Logged on theledger first 2 Decide The Data AccessCommitteeapproves,modifies orrefuses; noapproval, noaccess 3 Compute Atlas runs insidethe environment;Phylax admits it,debits the budget,checks the result 4 Answer An aggregate,non-identifyingresult with aprovenancecertificate and nodata. 5 Share The fee is paid; adefined sharegoes to the Statethrough thesponsoringinstitution 6 Withdraw A participantwithdrawsconsent: thecertificate isrevoked and everysystemregenerates A research request, step by step. A partner asks whether a variant common in the population changes the response to a widely prescribed medicine.
Why cohorts like those in Pakistan are important

Human knockouts, at national scale.

Homozygous loss-of-function variants are naturally occurring human knockouts. They are among the most valuable starting points in drug-target validation. They show what happens when a gene is switched off in a living person. They occur far more often in consanguineous populations. A governed national cohort, analysed with population-calibrated methods and interpreted with provenance, is a unique resource. Nothing like it exists anywhere else at national scale.

Life-sciences AI

The same evidence platform, applied to drug development.

AI strategy and regulatory-intelligence work for a clinical-stage cardiometabolic company; a delivered operational application for a UK engineering client. Under confidentiality; references on request.

Ready to explore governed research access?

Tell us the target class or the question. We will tell you what a cohort can and cannot answer, and on what terms.