Ensemble of Linear Models for Predicting Drug Properties∗

Abstract We propose a new classification method for prediction of drug properties, called the Random Feature Subset Boosting for Linear Discriminant Analysis (LDA). The main novelty of this method is the ability to overcome the problems with constructing ensembles of linear discriminant models based on generalised eignevectors of covariance matrices. Such linear models are popular in building classification-based structure–activity relationships. Introduction of ensembles of LDA models allows for analysis of more complex problems than by using single LDA, e.g. those involving multiple mechanisms of action. Using four datasets, we show experimentally that the method is competitive with other recently studied chemoinformatic methods, including support vector machines and models based on decision trees. We present an easy scheme for interpreting the model despite its apparent sophistication. We also outlin