Swarm-based support vector machine optimization for protein sequence-encoded prediction

Protein is considered the important macronutrient for most of the biochemical activities of all living organisms. Many healthcare applications involve protein to protein interactions (PPIs) to predict diseases, DNA characters, and more. PPI involves the mutation rate of amino acids, hydrophilicity, and hydrophobicity. The features are encoded with the Bayesian optimization (BO) based support vector machine (SVM approach), and the proteins are predicted with the proposed novel w-SVM which is the combination of the chameleon swarm algorithm (CSA) and SVM. This proposed method effectively predicts the protein from the database. The collected datasets are undergone a feature extraction process. This involves 2DLDA for mutation rate-based amino acids extraction. Continuous wavelet transform (CWT) for the hydrophilicity-based feature and discrete wavelet transform (DWT) for the hydrophobicity-based feature extraction. Finally, the experimental analysis is carried out in the MATLAB tool, and compared the results with state-of-the-art works such as LSTM, DSSP, and ARIMA approaches. The experimental shows that our proposed approach effectively predicts the proteins more than the other approaches.