A choquet integral-based multi-class classifier and its applications on the prediction of membrane protein types

In this paper, a novel aggregator information-based strategy for predicting membrane proteins types is introduced. In particular, we propose a framework of five Choquet Integrals (one Choquet Integral for each protein type) that are specialized to compute the global score of each class of proteins. These global scores are obtained by the combination of the partial evaluations of several membrane protein features provided by different individual classifiers. To compute the fuzzy measures associated with each Choquet Integral, we use a new unsupervised method (International Journal of Intelligent Systems, January 2008) proposed in the literature in which the concept of importance of attributes (in our case, the importance of the subsets of the classifiers) is replaced by that of information content in the subsets of classifiers. The parameters of the individual classifiers are adjusted with a conventional training dataset of 2059 sequences of aminoacids where 435 are Type I, 152 Type II, 1311 are multipass trans-membrane, 51 lipid-chain-anchored and 110 GPI-anchored type. The results obtained in this experiment, shows that our proposed method obtains a higher classification accuracy compared with the results obtained for several methods cited in the literature.