Machine Learning for Drug Design

A common step in drug design is the formation of a quantitative structure-activity relationship (QSAR) to model an exploratory series of compounds. A QSAR generalizes how the structure of a compound relates to its biological activity. There is growing interest in the application of machine learning techniques in QSAR modeling research. However, no single technique can claim to be uniformly superior to any other. This study introduced the ensemble machine learning, a set of classifiers whose individual decisions are combined in some way (typically by weighted or unweighted voting) to improve the performance of the overall system. A comparative study was carried out of two popular ensemble learning algorithms, Bagging and AdaBoost, for QSAR modeling. Two test case problems were studied: the inhibition of Escherichia coli dihydrofolate reductase (DHFR) by pyrimidines, and the inhibition of rat/mouse tumor DHFR by triazines. It was observed that the ensemble learning algorithms, Bagging and AdaBoost, can significantly improve the performance of Decision Tree C4.5 and 1-R (p<0.05), while Naïve Bayesian and 1-Nearest Neighbor did not benefit from ensemble learning. Furthermore, in general, AdaBoost outperformed Bagging on the tested data sets. Keywords-machine learning, drug design, QSAR, enselmble learning