Improving HIV-1 Integrase Inhibitors Prediction Using Hybrid Differential Evolution-Binary Particle Swarm Algorithm THESIS SUBMITTED IN PARTIAL FULLFILLMENT OF THE REQUIREMENTS FOR THE DEGREE MASTER OF SCIENCES IN COMPUTER SCIENCE

Due to the mutation of HIV enzymes and their resistance against current drugs, drug companies are exploring ways to develop stronger mutant resistant drugs. Dimeric aryl diketo acids have proven to be effective inhibitors to the HIV strand transfer mechanism of HIV-integrase. In order to create the best drug to fight HIV-integrase, it is important to know which features of the diketo acids have the biggest impact on reducing HIV enzyme activity. The use of evolutionary algorithms and predictive models, such as the differential evolutionary – binary particle swarm optimization (DE-BPSO) algorithm and Random Forest used in this research, can help find a small subset of the diketo acid’s chemical descriptors that are best able to predict the reduction in HIV enzyme activity. In this research the development of the DE-BPSO/MLR model is discussed and compared with the results against linear models tested in previous work. Also, through implementing Random Forest and TreeNet the results of linear models are compared with Non Linear models. Comparing both models will help determine which QSAR model yields the most effective HIV-IN inhibitors. Due to the nature of Acquired Immunodeficiency Syndrome (AIDS) and the current treatments’ numerous side effects and rapid emergence of drug resistant variants, maximizing the effectiveness of integrase inhibitors is an important milestone to current HIV research. Keyword: Data Mining, QSAR, RF, DE-BPSO, and HIV-Integres