prediction

Abstract. Several machine learning techniques, like neural networks, non-linear support vector machines and decision trees, have been used to model the specificity of HIV-1 protease and to extract specific patterns from peptides cleaved by this protease. Despite many studies, no perfect rules are already known to determine the cleavage of a peptide by HIV-1 protease. These rules are useful for designing specific and efficient HIV inhibitors. Our results show that the technique of mining association rules can find several specificity rules of HIV-1 protease which presents 100 % of cleavage probability. Recent papers on this subject show results in which the best rules present cleavage probability ranging from 16 % to 91%. 1