Michael S. Lee

This article presents a comparative analysis of two replica-exchange simulation methods for the structure refinement of protein loop conformations, starting from low-resolution predictions. The methods are self-guided Langevin dynamics (SGLD) and molecular dynamics (MD) with a Nosé-Hoover thermostat. We investigated a small dataset of 8- and 12-residue loops, with the shorter loops placed initially from a coarse-grained lattice model and the longer loops from an enumeration assembly method (the Loopy program). The CHARMM22 + CMAP force field with a generalized Born implicit solvent model (molecular-surface parameterized GBSW2) was used to explore conformational space. We also assessed two empirical scoring methods to detect nativelike conformations from decoys: the all-atom distance-scaled ideal-gas reference state (DFIRE-AA) statistical potential and the Rosetta energy function. Among the eight-residue loop targets, SGLD out performed MD in all cases, with a median of 0.48 Å reduction in global root-mean-square deviation (RMSD) of the loop backbone coordinates from the native structure. Among the more challenging 12-residue loop targets, SGLD improved the prediction accuracy over MD by a median of 1.31 Å, representing a substantial improvement. The overall median RMSD for SGLD simulations of 12-residue loops was 0.91 Å, yielding refinement of a median 2.70 Å from initial loop placement. Results from DFIRE-AA and the Rosetta model applied to rescoring conformations failed to improve the overall detection calculated from the CHARMM force field. We illustrate the advantage of SGLD over the MD simulation model by presenting potential-energy landscapes for several loop predictions. Our results demonstrate that SGLD significantly outperforms traditional MD in the generation and populating of nativelike loop conformations and that the CHARMM force field performs comparably to other empirical force fields in identifying these conformations from the resulting ensembles.

Computational sequence design methods are used to engineer proteins with desired properties such as increased thermal stability and novel function. In addition, these algorithms can be used to identify an envelope of sequences that may be compatible with a particular protein fold topology. In this regard, we hypothesized that sequence-property prediction, specifically secondary structure, could be significantly enhanced by using a large database of computationally designed sequences. We performed a large-scale test of this hypothesis with 6511 diverse protein domains and 50 designed sequences per domain. After analysis of the inherent accuracy of the designed sequences database, we realized that it was necessary to put constraints on what fraction of the native sequence should be allowed to change. With mutational constraints, accuracy was improved vs. no constraints, but the diversity of designed sequences, and hence effective size of the database, was moderately reduced. Overall, the best three-state prediction accuracy (Q₃) that we achieved was nearly a percentage point improved over using a natural sequence database alone, well below the theoretical possibility for improvement of 8–10 percentage points. Furthermore, our nascent method was used to augment the state-of-the-art PSIPRED program by a percentage point.

A comparative analysis is provided of the effect of different solvent models on the calculation of a potential of mean force (PMF) for determining the absolute binding affinity of the small molecule inhibitor pteroic acid bound to ricin toxin A-chain (RTA). Solvent models include the distance-dependent dielectric constant, several different generalized Born (GB) approximations, and a hybrid explicit/GB-based implicit solvent model. We found that the simpler approximation of dielectric screening and a GB model, with Born radii fitted to a switching-window dielectric-boundary surface Poisson solvent model, severely overpredicted the binding affinity as compared to the experimental value, estimated to range from −4.4 to −6.0 kcal/mol. In contrast, GB models that are parametrized to fit the Lee-Richards molecular surface performed much better, predicting binding free energy within 1−3 kcal/mol of experimental estimates. However, the predicted free-energy profiles of these GB models displayed alternative binding modes not observed in the crystal structure. Finally, the most rigorous and computationally costly approach in this work, which used a hybrid explicit/implicit solvent model, correctly determined a binding funnel in the PMF near the crystallographic bound state and predicted an absolute binding affinity that was 2 kcal/mol more favorable than the estimated experimental binding affinity.

Implicit solvent methods have become popular tools in the field of protein dynamics simulations, yet evaluation of their validity has been primarily limited to comparisons with experimental and theoretical data for small molecules. In this paper, we use a recently developed hybrid explicit/implicit solvent methodology to evaluate the accuracy of several Poisson-based implicit solvent models. Specifically, we focus on the calculation of electrostatic solvation free energies of various fixed conformations for two proteins. We show that, among various dielectric boundary definitions, the Lee−Richards molecular surface has the best agreement with hybrid solvent results. Furthermore, certain modifications of the molecular surface Poisson protocol provide varied results. For instance, simple modifications of atomic radii on charged residues generally improve absolute errors but do not significantly reduce relative errors among conformations. On the other hand, using a water-probe radius of 1.0 Å, as opposed to the standard value of 1.4 Å, to generate the molecular surface, moderately improves both absolute and relative results.

We present a new hybrid explicit/implicit solvent method for dynamics simulations of macromolecular systems. The method models explicitly the hydration of the solute by either a layer or sphere of water molecules, and the generalized Born (GB) theory is used to treat the bulk continuum solvent outside the explicit simulation volume. To reduce the computational cost, we implemented a multigrid method for evaluating the pairwise electrostatic and GB terms. It is shown that for typical ion and protein simulations our method achieves similar equilibrium and dynamical observables as the conventional particle mesh Ewald (PME) method. Simulation timings are reported, which indicate that the hybrid method is much faster than PME, primarily due to a significant reduction in the number of explicit water molecules required to model hydration effects. © 2004 Wiley Periodicals, Inc. J Comput Chem 25: 1967–1978, 2004

In this work, we explore the question of whether pK(a) calculations based on a microscopic description of the protein and a macroscopic description of the solvent can be implemented to examine conformationally dependent proton shifts in proteins. To this end, we introduce a new method for performing constant-pH molecular dynamics (PHMD) simulations utilizing the generalized Born implicit solvent model. This approach employs an extended Hamiltonian in which continuous titration coordinates propagate simultaneously with the atomic motions of the system. The values adopted by these coordinates are modulated by potentials of mean force of isolated titratable model groups and the pH to control the proton occupation at particular sites in the polypeptide. Our results for four different proteins yield an absolute average error of approximately 1.6 pK units, and point to the role that thermally driven relaxation of the protein environment in the vicinity of titrating groups plays in modulating the local pK(a), thereby influencing the observed pK1/2 values. While the accuracy of our method is not yet equivalent to methods that obtain pK1/2 values through the ad hoc scaling of electrostatics, the present approach and constant pH methods in general provide a useful framework for studying pH-dependent phenomena. Further work to improve our model to approach quantitative agreement with experiment is outlined.