Cooperative versus key residues in globular protein folding: an artificial neural network approach

Neural networks have been applied with success for protein sequence analysis. Besides, the protein folding process involves many scientific aspects that are not completely understood yet. We present two new artificial neural network approaches (forward and reverse) focusing on the capability of each amino acid to cause a specific folding-either by itself or with the cooperativity of other residues that are close in the primary structure. The forward approach looks for an association between the protein primary structure fragments and its folded structure; this approach should retain the cooperative features of residues. The reverse approach looks for an association between a fragment of protein tertiary structure and the amino acid placed in the center of the fragment; this approach should retain the singular influence that a key residue by itself has on that folding. The results obtained emphasize the cooperative nature of regular structures and the specific role that some residues plays in the folded structures.