Mutating HIV Protease Protein Using Ant Colony Optimization and Fuzzy Cognitive Maps: Drug Susceptibility Analysis

Understanding the dynamics of the resistance mechanisms in HIV proteins mutations is a key for optimizing the use of existing antiviral drugs and developing new ones. Several statistical and machine learning techniques have been proposed for predicting the resistance of a mutation to a certain drug using its genotype information. However, the knowledge publicly available for this kind of processing is majorly about resistant sequences, leading to highly imbalanced knowledge bases, which is a serious problem in classification tasks. In previous works, the authors proposed a methodology for modeling an HIV protein as a dynamic system through Fuzzy Cognitive Maps. The adjusted maps obtained not just allow discovering relevant knowledge in the causality among the protein positions and the resistant, but also achieved very competitive performance in terms classification accuracy. Based on these works, in this paper we propose an Ant Colony Optimization based method for generating possible susceptible mutations using the adjusted maps and biological heuristic knowledge. As a result, the mutations obtained allow drug experts to have more information of the behavior of the protease protein whenever a susceptible mutation takes place.