Ab initio construction of polypeptide fragments: Accuracy of loop decoy discrimination by an all‐atom statistical potential and the AMBER force field with the Generalized Born solvation model

Abstract

We describe a novel method to generate ensembles of conformations of the main-chain atoms {N, C␣, C, O, C␤} for a sequence of amino acids within the context of a fixed protein framework. Each conformation satisfies fundamental stereochemical restraints such as idealized geometry, favorable / angles, and excluded volume. The ensembles include conformations both near and far from the native structure. Algorithms for effective conformational sampling and constant time overlap detection permit the generation of thousands of distinct conformations in minutes. Unlike previous approaches, our method samples dihedral angles from fine-grained / state sets, which we demonstrate is superior to exhaustive enumeration from coarse / sets. Applied to a large set of loop structures, our method samples consistently near-native conformations, averaging 0.4, 1.1, and 2.2 Å mainchain root-mean-square deviations for four, eight, and twelve residue long loops, respectively. The ensembles make ideal decoy sets to assess the discriminatory power of a selection method. Using these decoy sets, we conclude that quality of anchor geometry cannot reliably identify near-native conformations, though the selection results are comparable to previous loop prediction methods. In a subsequent study (

A hierarchical approach to all‐atom protein loop prediction

“…This finding suggests that the use of statistical database potentials, as opposed to realistic models based on physical chemical principles, has a fundamental limitation in achieving high-resolution structural refinement. Other published results using such potentials to carry out loop predictions led to similar conclusions; in particular, the recent results of de Bakker et al 23 have also suggested that all-atom force fields are capable of qualitative improvements in accuracy relative to statistical potentials. It is difficult to see how the fine details of flexible loop structure can be captured by a potential function that in essence ignores solvation effects and electrostatics and incorporates conformational analysis only at a very approximate level.…”

“…Some studies have used purely gas-phase energetics, 18 whereas the early study of Moult and James 6 used an image-charge method to represent charge screening, and several studies have used simple distance-dependent dielectric models. 13 23 used a similar energy function on a large database of loop decoys. The energy function used here is similar (OPLS all-atom force field 24 -26 and Surface Generalized Born model of solvation 27,28 ), but the sampling algorithms used are entirely new and make it possible to apply a realistic, well-validated solvent model to a large test set of loops, with modest computational expense.…”

“…Xiang et al investigated all loops within their chosen protein data set; thus, the number of loops varies inversely with loop length due to natural abundance. Our results are reported as "global" RMSDs, obtained by superimposing the body of the protein; this choice is the same as that adopted by Xiang et al 8 and de Bakker et al 23 but differs from the "local" RMSD measure used by Fiser et al, 9 which provides no information about the orientation of the loop with respect to the body of the protein. If the objective is to use the structures for understanding interactions relevant to biological function and medicinal chemistry (e.g., docking candidate inhibitors), then accurate placement of the atoms of the loop in the context of the protein is of critical importance, and RMSD computed via loop stem superposition provides a reasonable measure of the effectiveness of the algorithm in accomplishing this task (ultimately, more direct measures, such as accuracy of docking into the resulting structure, would be even more relevant and insightful).…”

Membrane protein prediction methods

“…On the other hand, de novo/ab initio predictions for loop regions shorter than 12 residues tend to be reliable (113 , 114 ), if constructed using reasonable physicalchemical force fields that incorporate descriptions of the aqueous solution rather than a vacuum environment (115 , 116 ). Existing loop-prediction programs include Loopy (117), RAPPER (114) and PLOP (113) (Table 5); of these, PLOP is reportedly the most accurate, although it requires the most computational time, whereas Loopy is the fastest. Other strategies have been reported for loop modeling in GPCRs, including simulated annealing with Monte Carlo (116) and coarse-grained backbone dihedral sampling (118), although larger benchmarks are required to confirm their general applicability.…”

Analysis of the interaction between human steroid 21-hydroxylase and various monoclonal antibodies using comparative structural modelling

“…Alignments were obtained using FUGUE (19). The MAb models were built using MODELLER (30) and RAPPER (31,32) and validated with PROCHECK (25) and VERIFY3D (26). HBPLUS was used for hydrogen bond definition (33).…”