Research on Structure-Activity Relationship of HIV–1 Protease Based on Support Vector Machine Method

Abstract

Human immunodeficiency virus type 1 (HIV-1) is a kind of retrovirus and can cause AIDS (Acquired Immune Deficiency Syndrome). HIV-1 protease is a retroviral aspartyl protease which is essential for the life-cycle of HIV. HIV-1 protease inhibitors can decrease the activity of HIV-1, and this can reduce the infections of AIDS. In the study of HIV-1protease structure-activity relationship, we built a data set of 1630 compounds including 830 protease inhibitors and 800 HIV-1 protease decoys. Support Vector Machine (SVM) was used to build the classification model for HIV-1 protease inhibitors and decoys and the quantitative prediction model for HIV-1 protease inhibitors. In the study we use ADRIANA.Code software to calculate the descriptors of HIV-1 protease inhibitors and HIV-1 protease decoys, including 2D and 3D descriptors. The classification model and quantitative prediction model were built for HIV-1 protease inhibitors. The accuracy rates of classification models are over 98%, and Matthews Correlation Coefficient (MCC) of classification models are over 0.96. The linear regression coefficients of the quantitative prediction models are above 0.75