Complete populations of virtual patients for in silico clinical trials.

摘要: MOTIVATION Model-based approaches to safety and efficacy assessment of pharmacological drugs, treatment strategies, or medical devices (In Silico Clinical Trial, ISCT) aim decrease time cost for the needed experimentations, reduce animal human testing, enable precision medicine. Unfortunately, in presence non-identifiable models (e.g., reaction networks), parameter estimation is not enough generate complete populations Virtual Patient (VPs), i.e., guaranteed show entire spectrum model behaviours (phenotypes), thus ensuring representativeness trial. RESULTS We present methods software based on global search driven by statistical checking that, starting from a (non-identifiable) quantitative physiology (plus drugs PK/PD) suitable biological knowledge elicited experts, compute population VPs whose are representative whole phenotypes entailed (completeness) pairwise distinguishable according user-provided criteria. This enables full granularity control size employ an ISCT, guaranteeing while avoiding over-representation behaviours.We proved effectiveness our algorithm ODE-based female Hypothalamic-Pituitary-Gonadal axis, generating 4 830 264 stratified into 7 levels (at different behaviours), assessed its against 86 retrospective health records Pfizer, Hannover Medical School University Hospital Lausanne. The datasets respectively covered within Average Normalised Mean Absolute Error 15%, 20%, 35% (90% latter dataset 20% error).