AI digital twins and trial simulations for clinical development teams
Unlearn helps biopharma sponsors plan and analyze clinical trials using AI-generated digital twins, simulation, and statistical methods. A digital twin forecasts an individual participant’s clinical outcomes under control or standard of care, using their baseline data.
Teams can work with Unlearn before protocol finalization, during an ongoing study, or after readout. Its technology and scientific support help evaluate design assumptions, improve the precision of analyses, and inform subsequent development decisions.
These forecasts are modeled research inputs, not observed outcomes or an individual diagnosis. Biostatistics and clinical teams need to assess the relevant indication, baseline data, assumptions, and pre-specified analysis plan. Potential changes to power, sample size, or sensitivity depend on the study; they do not guarantee treatment efficacy, regulatory acceptance, or a faster trial.
Use models trained on historical patient-level clinical data to forecast progression under control or standard of care from baseline data. Evaluate the model and underlying dataset for the indication and study before using its predictions.
Structures prior-trial evidence into sourced parameters and compares scenarios across eligibility criteria, endpoints, comparators, and sample sizes. Solution Scientists help teams examine assumptions and trace claims to evidence.
PROCOVA uses prognostic information in covariate-adjusted trial analyses to support sample-size and power tradeoffs. The suitability and potential precision benefit depend on the model, endpoint, trial design, and statistical assumptions.
Unlearn scientists generate twins from collected baseline data, draft pre-specified analyses for the statistical analysis plan, and deliver adjusted analyses at interim and final readout.
Sponsors can build and deploy custom digital twin generators using proprietary data in their own environments. The technology can be delivered through a web application or within sponsor-controlled cloud infrastructure; scope and data handling should be agreed for the deployment.
Compare proposed populations, endpoints, comparators, and enrollment targets before protocol lock. Use simulations and sourced assumptions to document tradeoffs for governance review, without treating a prior sponsor’s savings as a guarantee.
Re-run design scenarios after new publications, competitor results, or blinded data change assumptions. Use pre-specified adjusted analyses for professional interim review; eligibility to change a study depends on its analysis and governance plan.
When randomization is not an option, assess model suitability and external-comparator assumptions for a single-arm or open-label study. A modeled comparator is not automatically interchangeable with observed randomized controls.
Re-analyze collected data to examine endpoint sensitivity, population characteristics, and variability, then use those findings to decide whether and how to design the next study.
The disease-specific models use a participant’s baseline data to forecast subsequent outcomes under control or standard of care.
Pre-specified digital twin analyses can be added to the statistical analysis plan before unblinding. Scenarios can also be revisited when evidence changes, and completed data can be re-analyzed after readout. The study team must assess the statistical and governance conditions for each use.
Request model specification sheets, review the indication and underlying data, and work with clinical and biostatistics specialists to assess baseline coverage, endpoints, assumptions, and analysis suitability. A forecast is not a guarantee of an individual outcome.
No. The product supports clinical-trial design and analysis. Digital twins forecast research outcomes under control or standard of care; they do not establish an individual diagnosis or direct patient treatment. Clinical and statistical teams remain responsible for interpreting the evidence.
Os dados de tráfego são apenas para referência.
