Abstract
Design of an efficient fingerprint that detects homologous proteins at distant sequence identity has been a great challenge. This paper proposes a strategy to extract an ideal-like fingerprint with high specificity and sensitivity from a group of sequences related to a fold. The approach is devised based on the assumptions that the critical residues for a protein fold may be conserved in three aspects, i.e. sequence, structure, and intramolecular interaction, and embedded in secondary structures. We hypothesized that the residues satisfying such conditions simultaneously may work as an efficient fingerprint. This idea was tested on protein folds of various classes, such as beta-strand rich, alpha + beta proteins and alpha/beta proteins with discrete sequence similarities. The fingerprint for each fold was generated by selecting the overlapped conserved residues (OCR) from the conserved residues obtained using independent three alignment methods, i.e. multiple sequence alignment, structure-based alignment, and alignment based on the interstrand hydrogen-bonds. The OCR fingerprints showed more than 90% detection efficiency for all the folds tested and were identified to be almost the minimal fingerprints composed of only critical residues. This study is expected to provide an important conceptual improvement in the identification or design of ideal fingerprints for a protein fold.