Optimal Length of Fragments for use in Protein Structure Prediction

Abstract

Proteins are three-dimensional structures that carry out many of the vital biological functions in organisms. Because structure, not amino acid sequence order, carry out certain functions, it is important to understand how proteins fold. Computational methods for protein structure prediction mentioned in the literature are computationally demanding. To reduce computational demand fragment libraries were introduced. Fragment libraries work by taking short segments of the polypeptide chain and limiting the amount of conformations that will be considered for a particular segment. In this paper an extensive analysis towards finding the optimal length of fragments contained within fragment libraries was conducted. An extensive analysis was done on protein structures stored in a ORDBMS to exploit its power. Experiments focused on the structural similarity between fragments of identical primary protein sub-sequence within different proteins, and amount of occurrences of similar or closely similar fragments within different proteins. Experimental results indicate that short to medium sized fragments have stronger structural correlations with matching fragments within different proteins.