Improving Pharmacological Research of HIV-1 Integrase Inhibition Using Differential Evolution - Binary Particle Swarm Optimization and Nonlinear Adaptive Boosting Random Forest Regression

In this work, we present results produced from a nonlinear QSAR model developed and implemented using evolutionary computation and Random Forest Regression to study the effectiveness of dimeric Aryl ß-Diketo Acids on HIV-1 Integrase enzyme inhibition. Dimeric Aryl ß-Diketo Acids have been proven to be effective inhibitors of the biological mechanism of protein transfer known as HIV-integrase. This research extends a previous study of Aryl ß-Diketo Acids for HIV-1 Integrase inhibition [1] that used linear QSAR models implemented using a Multiple Linear Regression (MLR) machine learning strategy and a hybridized Differential Evolution-Binary Particle Swarm Optimization (DE-BPSO) algorithm to select and identify, respectively, drug descriptors having the greatest inhibitory effect on HIV-1 Integrase. This comparative study uses a non-linear Random Forest Regression (RFR) strategy with adaptive boosting (AdaBoost) to generate QSAR models with greater predictive qualities in identifying optimal drug feature descriptors that can more effectively inhibit HIV protein enzyme activity.