A Practical <i>In Silico</i> Method for\nPredicting Compound Brain Concentration–Time Profiles: Combination\nof PK Modeling and Machine Learning

Given\nthe aging populations in advanced countries globally, many\npharmaceutical companies have focused on developing central nervous\nsystem (CNS) drugs. However, due to the blood–brain barrier,\ndrugs do not easily reach the target area in the brain. Although conventional\nscreening methods for drug discovery involve the measurement of (unbound\nfraction of drug) brain-to-plasma partition coefficients, it is difficult\nto consider nonequilibrium between plasma and brain compound concentration–time\nprofiles. To truly understand the pharmacokinetics/pharmacodynamics\nof CNS drugs, compound concentration–time profiles in the brain\nare necessary; however, such analyses are costly and time-consuming\nand require a significant number of animals. Therefore, in this study,\nwe attempted to develop an <i>in silico</i> prediction method\nthat does not require a large amount of experimental data by combining\nmodeling and simulation (M&S) with machine learning (ML). First,\nwe constructed a hybrid model linking plasma concentration–time\nprofile to the brain compartment that takes into account the transit\ntime and brain distribution of each compound. Using mouse plasma and\nbrain time experimental values for 103 compounds, we determined the\nbrain kinetic parameters of the hybrid model for each compound; this\ncase was defined as scenario I (a positive control experiment) and\nincluded the full brain concentration–time profile data. Next,\nwe built an ML model using chemical structure descriptors as explanatory\nvariables and rate parameters as the target variable, and we then\ninput the predicted values from 5-fold cross-validation (CV) into\nthe hybrid model; this case was defined as scenario II, in which no\nbrain compound concentration–time profile data exist. Finally,\nfor scenario III, assuming that the brain concentration is obtained\nat only one time point, we used the brain kinetic parameters from\nthe result of the 5-fold CV in scenario II as the initial values for\nthe hybrid model and performed parameter refitting against the observed\nbrain concentration at that time point. As a result, the RMSE/R2-values\nof the brain compound concentration–time profiles over time\nwere 0.445/0.517 in scenario II and 0.246/0.805 in scenario III, indicating\nthe method provides high accuracy and suggesting that it is a practical\nmethod for predicting brain compound concentration–time profiles.