Protein Closed Loop Prediction from Contact Probabilities

Abstract

According to the theory of Trifonov [1] a subset of closed loops found in globular proteins, descriptively named loop-n-lock (LNL) structures, are returns of a polypeptide chain trajectory to close contact with itself. These closed loops are distinguished by a characteristic length, 20 to 35 residues, and a ”lock” that stabilizes the loop. Evidence supports their contention that globular proteins are composed of linear combinations of such closed loops. Occupying an intermediate position in protein structure hierarchy i.e., between secondary structure and domain, makes these loops good candidates for units of a hierarchical folding model.

In this study we investigated the potential utility of such closed loops in protein structure prediction. We first proposed a method to predict closed loops using sequence information, specifically interaction potential and secondary structure potential. As far as we know, this is the first program to predict closed loops solely based on sequence information. Then to support their use in a hierarchical folding model, we explored the placement of secondary structure elements with respect to closed loops. Our investigations showed that 80% of secondary structures do not cross a closed loop boundary, lending support to the hypothesis that the closed loop is a higher level intermediate structure of proteins. Moreover, we explored the relationship between closed loops and aligned segments from protein chain pairs from the Dali data set . The results showed protein pairs which exhibited significant structural similarity were more likely to have conserved closed loops.