A multi-agent system for supporting the prediction of protein structures

In the last years numerous tools for predicting the three-dimensional structure of proteins have been proposed. Although their performance is constantly increasing, they are not sufficiently general to be exploited in any prediction problem. As a consequence, in order to improve the prediction quality, it could be particularly useful to jointly apply different prediction tools to the same problem and, then, integrate their results. In such a context, since the various predictors could have different performances on the same prediction, the choice of the predictors to jointly apply for guaranteeing the best performances appears crucial. In this paper we propose X-MACoP, an XML multi-agent system for supporting a user in the prediction of the three-dimensional structures of proteins. In particular, X-MACoP carries out the following tasks, in a way completely transparent for the user: (i) choice of the most promising predictor team to apply for the prediction problem of interest for the user; (ii) integration of the results produced by the predictors of the team for constructing a unique global prediction for the user; (iii) possible translation of predictor inputs and outputs in such a way that a user handles a unique data format.