摘要
HIV-1 protease is a crucial enzyme in the HIV life cycle and serves to cleave the polyprotein that is involved in the formation of mature viruses. Due to its sensitivity and important function in virion creation, predicting cleavage classification can aid in the development of HIV-1 protease inhibitors that can improve antiretroviral therapy. Currently available methods are less effective at maintaining high prediction accuracy and consistency when applied to data with class bias and can be simplified and optimized. A prediction method that focused solely on sequential data was proposed and used in this study. A simplified swarm optimization (SSO) algorithm was applied to classifying HIV-1 protease cleavage data, which consisted of octamers that would be cleaved between the fourth and fifth amino acid, and orthogonal array testing was incorporated to improve efficiency. The prediction accuracy was assessed by applying the SSO algorithm to datasets found in the UCI Machine Learning Repository. Our experimental results show that SSO is an effective predictor of HIV-1 protease cleavage, exhibiting prediction accuracy that compared favorably to existing methods for both data with little class bias and data that contains class bias. Additionally, our prediction accuracy results suggest that the use of physicochemical features, as opposed to solely sequential features, does not improve performance. © 2019 Association for Computing Machinery.