A procedure for evaluating models of structures with flexible loops

Friday, November 08, 2013 — Poster Session IV
2:00 p.m. – 4:00 p.m.	FAES Academic Center (Upper-Level Terrace)	NCI	COMPBIO-1

Authors

H. Bai
C.H. Tai
B. Lee

Abstract

Loop modeling is important for the prediction of protein structure because loops make up half of the residues in a protein on average and because they are often part of the active site of the protein. However, evaluating loop modeling is complicated because loops are often flexible and can assume multiple conformations. Here we propose a set of procedures for evaluating models for flexible loops and illustrate it using a set of three NMR structures of CASP10 (R0003, T0709 and T0711), which were engineered from the same protein by differently mutating one loop to change their binding specificity. Some of the results of evaluating predictions for these three targets are the following. (1) Even though the loops in the three target structures are flexible, there are measurable differences in the ensemble of conformations among the three. (2) Predicted models can be ranked and good models identified. (3) Good models are specific, i.e. they are closer to their intended target than to a wrong target. (4) A few top predicted models are indistinguishable from the target structures in the sense that they are as similar to the target ensemble as one of the target structure itself is.