The Platform
From multimodal cohort data to subgroup failure predictions
Valinor Discovery's platform ingests clinical and genomic data, builds probabilistic virtual patient cohorts, and simulates trial population outcomes across candidate compounds. All of this before a single participant is enrolled.
Data Ingestion
Designed for the data formats clinical development teams already produce
No bespoke data pipelines. No manual reformatting. The platform connects directly to the structured data outputs from your existing clinical data systems.
CDISC SDTM and ADaM
Structured trial datasets in standard CDISC format are ingested directly. Subject-level domains (DM, EX, LB, VS) and derived analysis datasets map into the virtual patient parameter space without transformation.
HL7 FHIR patient records
De-identified, aggregate-level FHIR bundles covering Patient, Condition, Observation, and MedicationRequest resources. We work from cohort-level conditional statistics, not individual patient records.
Genomic panel outputs
VCF files and targeted sequencing outputs from WES and gene panel assays. Variant allele frequencies and copy number variation summaries integrate into the biomarker axis conditioning for each virtual patient.
Lab time-series and phenotypic annotations
CBC, metabolic panel, and proteomic marker longitudinal data alongside ICD-10 coded phenotypic annotation files. Temporal conditioning captures the comorbidity and disease progression patterns that drive subgroup heterogeneity.
Simulation Engine
A Bayesian network at the center of every virtual patient
Each virtual patient is a statistical draw from a probabilistic model trained on real cohort data. The network encodes conditional dependencies between biomarker axes, comorbidities, and treatment histories. Structural relationships from the data, not assumptions.
Cohort statistics extraction
Aggregate-level statistics are extracted from the ingested datasets: means, variances, and joint frequency tables across biomarker and clinical axes. No individual patient record leaves your environment.
Bayesian network construction
A directed acyclic graph is fit to the extracted conditional distributions, incorporating literature-curated epidemiological priors for disease prevalence and comorbidity co-occurrence.
Virtual cohort sampling
1,000 to 10,000 virtual patients are sampled from the fitted network. The resulting synthetic cohort has the correct epidemiological shape for your development context without containing any real patient data.
Subgroup signal detection
Trial exposure conditions are simulated across the virtual cohort. The platform identifies which biomarker-defined subgroups show efficacy signal collapse at realistic enrichment thresholds. This happens before protocol lock, when the design can still change.
Platform Output
What the platform returns to your development team
Simulation results are delivered as structured reports, not raw model outputs. Each report is designed to map directly to a protocol design or target prioritization decision.
Synthetic cohort composition report
Cohort profile summary for 1,000 to 10,000 virtual patients: biomarker distributions, comorbidity prevalence, treatment history composition. Includes confidence intervals reflecting uncertainty in input data coverage.
Subgroup failure probability distributions
For each candidate compound and exposure condition, a probability distribution over subgroup-level efficacy outcomes. Identifies which biomarker-defined strata are likely to dilute overall trial response.
Enrichment criteria recommendations
Ranked enrichment strategies with predicted improvement in trial success probability, expressed as specific inclusion/exclusion criteria language for protocol review.
Target-population alignment score
A single summary metric per candidate compound expressing the concordance between the predicted responder subgroup and the intended trial population. Designed for direct integration into target prioritization decisions.
Integration
Fits into your existing translational workflow
Valinor Discovery is not a standalone analytics environment. The platform is designed to insert at specific decision points in Phase I-II development without requiring a separate data team workflow.
Technical Specifications
Platform specifications for development teams evaluating data readiness
Ready to evaluate data readiness for your program?
We work with a small number of translational and clinical development teams at a time. Tell us about your compound, your data, and the development question you are trying to answer before Phase II enrolment.
Request Early Access Explore the science