iMRSA-Fuse: A Fast and Accurate Computational Approach for Predicting Anti-MRSA Peptides by Fusing Multi-View Information

作     者：Charoenkwan, Phasit Schaduangrat, Nalini Moni, Mohammad Ali Shoombuatong, Watshara 

作者机构：Chiang Mai Univ Coll Arts Media & Technol Modern Management & Informat Technol Chiang Mai 50200 Thailand Mahidol Univ Fac Med Technol Ctr Res Innovat & Biomed Informat Bangkok 10700 Thailand Charles Sturt Univ AI & Cyber Futures Inst AI & Digital Hlth Technol Bathurst NSW 2795 Australia 

出 版 物：《IEEE TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS》 

年 卷 期：2025年第22卷第1期

页      面：2-12页

核心收录：

基　　金：National Research Council of Thailand Mahidol University [N42A660380] College of Arts, Media and Technology, Chiang Mai University, Chiang Mai University, Mahidol University 

主　　题：Peptides Amino acids Encoding Bioinformatics Biological system modeling Feature extraction Computational modeling Accuracy Random forests Immune system Anti-MRSA peptide sequence analysis bioinformatics machine learning feature representation multi-view feature 

摘      要：Methicillin-resistant S. aureus (MRSA) has prominently emerged among the recognized causes of community-acquired and hospital infections. We proposed a novel computational approach, iMRSA-Fuse, based on a multi-view feature fusion strategy for fast and accurate anti-MRSA peptide identification. In iMRSA-Fuse, we explored and integrated 12 different sequence-based feature descriptors from multiple perspectives, in conjunction with 12 popular machine learning (ML) algorithms, to construct multi-view features that were able to fully capture the useful information of anti-MRSA peptides. Additionally, we applied our customized genetic algorithm to determine a set of multi-view features to enhance its discriminative ability. Based on a series of comparative results, our multi-view features exhibited the most discriminative ability compared to several conventional feature descriptors. Moreover, concerning the independent test dataset, iMRSA-Fuse achieved the best balanced accuracy (BACC) and Matthew s correlation coefficient (MCC) of 0.997 and 0.981, respectively with an increase of 3.93 and 7.78%, respectively. Finally, to facilitate the large-scale identification of candidate anti-MRSA peptides, a user-friendly web server of the iMRSA-Fuse model is constructed and is freely accessible at https://***/iMRSA-Fuse. We anticipate that this new computational approach will be effectively applied to screen and prioritize candidate peptides that might exhibit the great anti-MRSA activities.