Effect of Molecular Descriptor Feature Selection in Support Vector Machine Classification of Pharmacokinetic and Toxicological Properties of Chemical Agents

Abstract

Statistical-learning methods have been developed for facilitating the prediction of pharmacokinetic and toxicological properties of chemical agents. These methods employ a variety of molecular descriptors to characterize structural and physicochemical properties of molecules. Some of these descriptors are specifically designed for the study of a particular type of properties or agents, and their use for other properties or agents might generate noise and affect the prediction accuracy of a statistical learning system. This work examines to what extent the reduction of this noise can improve the prediction accuracy of a statistical learning system. A feature selection method, recursive feature elimination (RFE), is used to automatically select molecular descriptors for support vector machines (SVM) prediction of P-glycoprotein substrates (P-gp), human intestinal absorption of molecules (HIA), and agents that cause torsades de pointes (TdP), a rare but serious side effect. RFE significantly reduces the number of descriptors for each of these properties thereby increasing the computational speed for their classification. The SVM prediction accuracies of P-gp and HIA are substantially increased and that of TdP remains unchanged by RFE. These prediction accuracies are comparable to those of earlier studies derived from a selective set of descriptors. Our study suggests that molecular feature selection is useful for improving the speed and, in some cases, the accuracy of statistical learning methods for the prediction of pharmacokinetic and toxicological properties of chemical agents.

The impact of training set data distributions for modelling of passive intestinal absorption

“…The choice of 50% was arbitrary although it has also been used in previous studies (Niwa 2003). There have been a number of different cutoffs used, from 10% (Palm et al 1997) up to 70% (Xue et al 2004), with no standard defined. Table 4, referring to models built from training set TS1, shows that for the classification of the validation set the best overall classification accuracy was 0.958 (481/502), the highest specificity value was 0.952 (441/460) and the best sensitivity was 0.959 (40/42), all using model 3.…”

Classification of oral bioavailability of drugs by machine learning approaches: a comparative study.

“…Area under ROC curve was found to be 0.967, whereas Youden's index was calculated as 0.84. Quite a few researchers have been tried to generate absorption models using different machine learning approaches and reported good results [6,24]. This is the first time we are presenting a comparative study between three potential machine learning approaches viz.…”

Classification of Highly Unbalanced CYP450 Data of Drugs Using Cost Sensitive Machine Learning Techniques.

“…59 The problem of selecting properties which are responsible for given outputs occurs in various machine learning applications. [60][61][62] We use feature selection methods with the objective to detect features that are responsible for the underlying class structure. In addition, we search for feature combinations that reflect or even outperform results using all features.…”