State-of-the-Art Genetic Programming for Predicting Human Oral Bioavailability of Drugs

Abstract

Being able to predict the human oral bioavailabilityfor a potential new drug is extremely important for thedrug discovery process. This problem has been addressedby several prediction tools, with Genetic Programmingproviding some of the best results ever achieved. Inthis paper we use the newest state-of-the-artdevelopments of Genetic Programming, in particular thelatest bloat control method, to find out exactly howmuch improvement we can achieve on this problem.We showexamples of some actual solutions and discuss theirquality from the practitioners' point of view,comparing them with previously published results. Weidentify some unexpected behaviors and discuss the wayfor further improving the practical usage of theGenetic Programming approach.