Validation of predicted conformal intervals for prediction of human clinical pharmacokinetics

IntroductionConformal prediction (CP) methodology sits on top of machine learning methods and produces prediction confidence intervals that depend on how "strange" (non-conforming) test compounds are compared to training set compounds. CP has previously been successfully applied for prediction of steady-state volume of distribution (Vss) in humans, with 69 % of observations within the prediction interval at a 70 % confidence level. We have developed CP models for a variety of human pharmacokinetic (PK) parameters and validated their predictive accuracy (predicted vs observed estimates), but not validated prediction confidence intervals for them. The main objective of this study was to predict 70 % confidence intervals for Vss, unbound fraction in plasma (fu), intrinsic metabolic clearance (CLint), fraction absorbed passively (fa,passive) and maximum fraction dissolved (fdiss) for a variety of compounds in man and investigate the consistency between prediction intervals and observed/measured values. MethodologyCP models featured in the ANDROMEDA software by Prosilico were used for prediction of 70 % confidence intervals of Vss, fu, CLint, fa,passive and fdiss for compounds from different chemical classes and with broad physicochemical variety and for small drugs marketed in 2021. Results70 % prediction confidence intervals for 217, 117, 117, 89 and 89 compounds were produced for Vss, fu, CLint, fa,passive and fdiss, respectively. 78 % (expected 70 %) of observed data were within 70 % confidence intervals for the parameters. 70 % of predictions of Vss, fu, CLint fa,passive and fdiss are expected to have errors of maximally 2-, 4- and 6-fold and 7 and 12 %, respectively, which is in line with prediction errors. These findings validate the CP methodology. ConclusionIn conclusion, the results further validate CP models and confidence intervals of ANDROMEDA for prediction of human PK.