Development and validation of machine learning classificatory models for predicting bioavailability of drugs in the central nervous system

JUAN FRANCISCO MORALES; MARÍA ESPERANZA RUIZ; ROBERT E. STRATFORD; ALAN TALEVI

Orlando, FL

Congreso; 10th. American Conference on Pharmacometrics; 2019

International Society of Pharmacometrics (ISoP)

Optimizing brain bioavailability is highly relevant during the development of drugs targeting central nervous system. In this work, we developed new in silico models to predict Kpuu,brain,ss, the unbound brain and plasma drug concentrations ratio at steady state conditions. Good practices of predictive QSAR modeling were followed for model development and different machine learning algorithms were tested: Support Vector Machine, Gradient Boosting, k-nearest neighbors, classificatory Partial Least Squares, Random Forest, Extreme Gradient Boosting and Linear Discriminant Analysis. The best performance was achieved by both gradient boosting methods, with an area under the ROC curve and accuracy in the test set equal or greater than 0.951 and 91.8%, respectively. Additional models were also trained by removing known P-gp and BCRP substrates from the original dataset. Since the experimental data used are provided, our modeling approaches may be reproduced, verified and even expanded, as a useful tool to assist drug discovery processes