A hierarchical approach to all‐atom protein loop prediction

Abstract

The application of all-atom force fields (and explicit or implicit solvent models) to protein homology-modeling tasks such as side-chain and loop prediction remains challenging both because of the expense of the individual energy calculations and because of the difficulty of sampling the rugged all-atom energy surface. Here we address this challenge for the problem of loop prediction through the development of numerous new algorithms, with an emphasis on multiscale and hierarchical techniques. As a first step in evaluating the performance of our loop prediction algorithm, we have applied it to the problem of reconstructing loops in native structures; we also explicitly include crystal packing to provide a fair comparison with crystal structures. In brief, large numbers of loops are generated by using a dihedral angle-based buildup procedure followed by iterative cycles of clustering, side-chain optimization, and complete energy minimization of selected loop structures. We evaluate this method by using the largest test set yet used for validation of a loop prediction method, with a total of 833 loops ranging from 4 to 12 residues in length. Average/median backbone rootmean-square deviations (RMSDs) to the native structures (superimposing the body of the protein, not the loop itself) are 0.42/0.24 Å for 5 residue loops, 1.00/0.44 Å for 8 residue loops, and 2.47/1.83 Å for 11 residue loops. Median RMSDs are substantially lower than the averages because of a small number of outliers; the causes of these failures are examined in some detail, and many can be attributed to errors in assignment of protonation states of titratable residues, omission of ligands from the simulation, and, in a few cases, probable errors in the experimentally determined structures. When these obvious problems in the data sets are filtered out, average RMSDs to the native structures improve to 0.43 Å for 5 residue loops, 0.84 Å for 8 residue loops, and 1.63 Å for 11 residue loops. In the vast majority of cases, the method locates energy minima that are lower than or equal to that of the minimized native loop, thus indicating that sampling rarely limits prediction accuracy. The overall results are, to our knowledge, the best reported to date, and we attribute this success to the combination of an accurate all-atom energy function, efficient methods for loop buildup and side-chain optimization, and, especially for the longer loops, the hierarchical refinement protocol.

Prediction of Protein Loop Conformations Using the AGBNP Implicit Solvent Model and Torsion Angle Sampling

“…Based on our analysis of their data, they had obtained ten energy errors and eight sampling errors. 26 In comparison, we find eleven energy and seven sampling errors with SGB/NP without crystal symmetry but we find only eight energy errors and five sampling errors with SGB/NP with crystal symmetry. This might indicate that crystal symmetry is important for prediction accuracy; however, we obtained two energy errors and five sampling errors using AGBNP+ without the presence of the crystal environment.…”

“…[48][49][50] We previously showed 44 that the OPLS-AA/AGBNP effective potential was able to consistently score native loop conformations more favorably than non-native decoy loop conformations generated by PLOP using the OPLS-AA/SGB/NP effective potential. 26 The present work extends that work by including a larger set of loops as well as longer loops targets, and by employing the OPLS-AA/AGBNP model directly in the conformational search and optimization procedure implemented in PLOP. We also evaluate various parameterizations of the AGBNP model to determine the role of the non-polar model and of the correction terms we developed aimed at reducing the occurrence of intramolecular ion pairing, and we compare them to the distance dependent dielectric and the Surface Generalized Born (SGB/NP) 51,52 solvation models as implemented in the PLOP program.…”

“…26. During loop build-up, a series of filters of increasing complexity is applied to eliminate unreasonable conformations as early as possible.…”

“…The standard 9-residue loop prediction scheme is based on the procedure described in detail by Jacobson et al 26 For loops which the standard version of loop prediction fails to find lowenergy, native-like conformations, we attempted to predict these loops with an extended version of the loop prediction algorithm. An extended version of this scheme involves using twice the number of clusters (from 36 to 72) and reduced ofac (overlap factor) coefficients (0.5 instead of 0.75) during the initial prediction stage.…”

Study of the interaction of Huperzia saururus Lycopodium alkaloids with the acetylcholinesterase enzyme

“…The coordinates of the missing residues (Pro259, Gly260, Gly261, Thr262 and Gly263) were added with the side chain prediction tools included in Maestro. [55] The protonation states of the amino acids in the protein were assigned with the H++ server, [56,57] and the ones of the residues in the catalytic gorge (His447, Glu334, Glu202 and Asp74) on the basis of previous representative work by McCammon et al [58] For A3 and A4.2, the search that provided the best exploration was found to be the combination of the genetic algorithm and a local search refinement (GA-LS). The search space was defined using AutoGrid, and the same grid was used for all the simulations with the three programs.…”