Bioactive molecule prediction using majority voting-based ensemble method

The current rise in the amount of data generated has necessitated the use of machine learning in the drug discovery process to increase productivity. It is therefore important to predict molecular compounds which are biologically active and capable of drug-target interaction. Various machine learning methods have been used in predicting bioactive molecular compounds in order to deal with the large volume of data being generated. This study investigates the Majority Voting ensemble method using different combinations of 5 commonly-used machine learning algorithms, including Support Vector Machine, Decision Tree, Naive Bayes, k-Nearest Neighbor, and Random Forest on three chemical datasets DS1, DS2, and DS3 which consist of structurally heterogeneous and homogeneous molecules and are commonly used in other studies. The results show that Majority Voting has a better performance, based on all the evaluation metrics used, compared to each of the machine learning algorithms as individual classifiers. It also shows the Majority Voting ensemble method as effective in the prediction of both heterogeneous and homogeneous bioactive molecular compounds, using statistical evaluation.