Minimal and optimal subsites set of HIV-1 protease cleavage site based on rough set

Identifying the cleavage sites in proteins by HIV protease will greatly expedite the pace in searching for proper inhibitors of HIV protease. In this study, we focus on how many subsites are needed for predicting the cleavage sites. Using the rough set theory of information gain and reduct to search for the minimal and optimal subsites set, we found that, instead of the 8-subsite octapeptide model as usually adopted, the 6-subsite hexapeptide model would suffice to obtain equally good or even slightly better results. To verify the conclusion, prediction with the 6-subsite hexapeptide mode and the 8-subsite octapeptide model are both studied. And the predictive rate and ROC curve are also reported. Such a finding not only suggests many tedious labors could be avoided in synthesizing peptide inhibitor candidates, but also implies it would be possible to design smaller peptide drugs based on the distorted key theory.