A machine learning tool for computer aided molecular design

Abstract

A description is given of KAMD (knowledge-aided molecular design), a machine learning tool that helps biochemists to reduce the number of experiments needed for molecular or, drug design processes. A report is presented on experience in applying machine learning techniques to a complex real-world problem. In this context, dynamic bias management is presented as a critical mechanism to deal with complex problems that typically exhibit a large number of distinct disjuncts. A summary of KAMD results for the system of thermolysin enzyme inhibitors is presented and discussed.<>