| ABSTRACT : |
In this paper, the biological activities of HIV-1 protease inhibitor compounds are predicted with Standard Fuzzy ARTMAP (FAM) and GA-FAMR, from the attributes describing the molecular descriptor of the compounds. Self-organized maps (SOM) have been applied to analyze the similarities of chemical compounds and to select from a given pool of molecular descriptors the smallest and more relevant subset needed to build robust QSAR (Quantitative Structure-Activity Relationship) models based on FAM. FAM is provided with 196 sets of data, out of which 176 are used for training and the remaining 20 are used for testing. The data are normalized and fed to the neuro-fuzzy network and the output indicates whether the compound is a suitable inhibitor for the HIV-1 virus. The analysis is done with and without Genetic Algorithm (GA) for the small dataset and GA-FAMR algorithm is used to optimize the relevance's assigned to the training data. The performance of the integrated SOM-FAM was evaluated and the prediction accuracy obtained is 93.09%.
Key words: Fuzzy ARTMAP, protease inhibitor, Self Organizing Map, Genetic Algorithm, QSAR |
|