LRC (logistic regression model and clustering approach)

A computational method based on a combination of physicochemical and structural properties to
predict the B-cell epitopes. We model the surface of antigen with a weighted graph, which vertices represent surface residues and edges denote the relation between residues. For each residue, the physicochemical and structural properties, as well as the probability of occurrence each of amino acid types in epitope regions are exploited. Then each vertex of graph is weighted based on logistic regression model and these properties. Also, we utilize the Markov cluster algorithm to cluster the weighted graph, then the clusters with less average weights are removed. Finally, the vertices with maximum weight into the candidate cluster are predicted as epitope residues.

Click here to download