Identification of non random motifs in loops using a structural alphabet

Loops in protein structures are irregular regions classically seen as "random". Some particular recurrent motifs have however been described (e.g. turns). Moreover, loop regions are highly variable both in sequence and in structure. Thus the prediction of loop conformation is a very actual and challenging problem. A structural alphabet allows the decomposition of 3D protein structures into short fragments. It is a useful tool to study the variety of loop conformations, allowing a very fast identification of common structural motifs. In this study HMM-SA, a structural alphabet based on hidden Markov model, is used to extract 36 over-represented recurrent structural motifs of length seven residues occurring in loops. These structural motifs are found in 22% of long loops (more than 13 residues). They are structurally very conserved, display particular sequence specificity and loop type/length preference. They cover classical motifs as turns but also novel fragments. The prediction of these particular motifs will be very useful to predict the conformation of long loops