Predicting protein-protein binding sites by a support vector machine approach

Identifying the interface between two interacting proteins is a crucial clue to the function of protein. We have combined a support vector machine (SVM) approach with PP_SITE, NACCESS to predict protein-protein binding sites. Using leave-one-out cross-validation procedure, we were able to successfully predict the location of binding sites on 98.09% (highest) and 61.31% (lowest). We achieved comparable success rates to the leave-one-out cross validation.