Computational Prediction of Influenza Neuraminidase Inhibitors Using Machine Learning Algorithms and Recursive Feature Elimination Method

Abstract

Listen

Translate article

Recent outbreaks of highly pathogenic influenza have highlighted the need to develop novel anti-influenza therapeutics. Neuraminidase has become the most important target for the treatment of influenza virus. In this study, classification models were developed from a large training dataset containing 457 neuraminidase inhibitors and 358 non-inhibitors using random forest and support vector machine algorithms. Recursive feature elimination (RFE) method was used to improve the accuracy of the models by selecting the most relevant molecular descriptors. The performances of the models were evaluated by five-fold cross-validation and independent validation. The accuracies of all the models are over 86% in both validation methods. This work suggests machine learning algorithms combined with RFE method can be used to build useful models for predicting influenza neuraminidase inhibitors.