Prediction of surface loops of protein-folds from multiple alignments of homologous sequences.

Multiple alignments of distantly related homologous sequences may be used for the construction of consensus sequences that identify conserved motifs, variable segments and regions that tolerate gap events. It is suggested that such consensus sequences may be used for the prediction of key features of protein-folds. The validity of the proposed approach is illustrated in the case of the alpha 2 mu globulin superfamily: the consensus sequence derived from the multiple alignment of sequences succeeded in identifying conserved structural motifs and in predicting the location of surface loops that connect these motifs.