Binary Gravitational Search Algorithm Feature Selection method for DTI

Anticipating the interplay between pharmaceuticals and proteins stands as an indispensablestride within the realm of drug advancement, leading to the discovery of innovative methods.Experimental approaches relying on clinical treatments to identify these relationships are timeconsuming,expensive, labor-intensive, and complex [۱,۲,۳]. New computational methods, particularlythose employing machine learning algorithms, offer a cost and time-efficient alternative to traditionalexperimental approaches[۴]. This paper presents an innovative computational framework designed toforecast drug-target interactions, where various features are extracted from a dataset consisting of drugsand their corresponding SMILES, and proteins with their protein sequences. To address the abundanceof features, BGSA (Binary Gravitational Search Algorithm) is applied. The carefully chosen featureshave the potential to significantly enhance the accuracy of the classification model, and minimizing thelearning process's computational cost[۵]. To improve the performance of BGSA, mutation operators areused to introduce random changes in the search process to avoid getting stuck in local optima. Suchactions facilitate the algorithm's exploration of uncharted space and improve the quality of the solutions[۶]. The Super Vector Machine (SVM) classification is utilized to predict the efficiency of the methodusing the ultimately chosen features. The SVM classifier's accuracy on the designated golden standarddatasets (Enzyme, ion channels, G-protein-coupled receptors, nuclear receptors) is provided belowbefore feature selection: ۹۵.۹۴%, ۹۵.۰۸%, ۸۷.۹۹%, and ۷۰.۸۳% and after feature selection are ۹۸.۶۳%,۹۸.۴۷%, ۹۶.۰۶%, and ۹۷.۲۲% respectively. According to the experimental results, the proposed methodshowcases a commendable level of accuracy in predicting drug-target interactions (DTI) andharmonizes effectively with methodologies advocated in existing publications.

کلیدواژه ها:

Drug-Target Interactions ، Feature Selection ، Binary Gravitational Search ، SVM