Predicting protein structure from long-range contacts

Abstract

Short-range and long-range contacts are important in forming protein structure. The proteins can be grouped into four different structural classes according to the content and topology of α-helices and β-strands, and there are all-α, all-β, α/β and α+β proteins. However, there is much difference in statistical property for those classes of proteins. In this paper, we will discuss protein structure in the view of the relative number of long-range (short-range) contacts for each residue. We find the percentage of residues having a large number of long-range contacts in protein is small in all-α class of proteins, and large in all-β class of proteins. However, the percentage of residues is almost the same in α/β and α+β classes of proteins. We calculate the percentage of residues having the number of long-range contacts greater than or equal to (≥) N L =5, and 7 for 428 proteins. The average percentage is 13.3%, 54.8%, 41.4% and 37.0% for all-α, all-β, α/β and α+β classes of proteins with N L =5, respectively. With N L increasing, the percentage decreases, especially for all-α class of proteins. In the meantime, the percentage of residues having the number of short-range contacts greater than or equal to N S (≥ N S ) in protein samples is large for all-α class of proteins, and small for all-β class of proteins, especially for large N S . We also investigate the ability of amino residues in forming a large number of long-range and short-range contacts. Cys, Val, Ile, Tyr, Trp and Phe can form a large number of long-range contacts easily, and Glu, Lys, Asp, Gln, Arg and Asn can form a large number of long-range contacts, but with difficulty. We also discuss the relative ability in forming short-range contacts for 20 amino residues. Comparison with Fauchere–Pliska hydrophobicity scale and the percentage of residues having large number of long-range contacts is also made. This investigation can provide some insights into the protein structure.