News and Publications

Molecular Dynamics of Proteins and Peptides

C.L. Brooks III, T. Cleveland, M.F. Crowley, B. Dominy, Z. Guo, J.D. Hirst, R.T. Morton, J. Radkiewicz, J.-E. Shea, F.B. Sheinerman, W.A. Shirley, M. Vieth, W.S. Young*

* Molecular Simulations Inc., San Diego, CA

Understanding the atomic-level forces that determine the structure of proteins, peptides, and protein-peptide complexes and the processes by which these structures are adopted is essential to complete knowledge of protein and peptide structure and function. To address such questions, we use statistical mechanics, molecular simulation, statistical modeling, and quantum chemistry.

Building atomic-level models to simulate biophysical processes (e.g., protein folding or ligand binding to a biological receptor) requires (1) the development of potential functions that accurately represent the atomic interactions and (2) the use of quantum chemistry to aid in characterizing these models. Calculation of thermodynamic properties requires the development and implementation of new theoretical and computational approaches that connect averages over atomistic descriptions to experimentally measurable thermodynamic and kinetic properties.

Interpreting experimental results at a more atomic level (e.g., to obtain a detailed description of structure from circular dichroism spectra or hydrogen exchange experiments on proteins and peptides) leads to the development of theoretical models for these processes. Additionally, the massive computational resources needed to produce atomic-level descriptions of proteins, peptides, and protein-peptide complexes in solution motivate efforts aimed at the efficient use of new computer architectures, including large supercomputers. Each of the tools or areas mentioned represents areas of development in our research program in computational biophysics.

LIGAND DOCKING

Using the theoretical techniques of computational chemistry to design potential inhibitors for known biological receptors requires (1) methods that allow accurate assessment of ligand affinity for the receptor of interest, relative to other ligands, and among different "binding modes" for the same ligand and (2) efficient search methods to explore and locate different ligand-receptor conformations. Such methods are elements of the computational techniques of ligand docking. We recently examined these 2 elements in the context of a database of ligand-receptor pairs of known structure.

We first addressed identification of suitable energy functions and "scoring methods" to assess these functions. We developed a scoring function to assess binding potentials and discriminate between energy functions used in computational docking studies. The scoring function is based on principles of energy-landscape analysis. It separates those functions that provide an accurate determination of binding affinities among families of correctly and incorrectly docked ligand conformations from those that do not. We detected key features of the energy functions that yielded the greatest discrimination. These included suitable screening of electrostatic interactions to mimic solvent influences on ligand-receptor interactions and sufficient detail in the energy function. For example, we found that grid-based energy functions often did not provide adequate discrimination between correctly docked and incorrectly docked ligands.

The nature of the energy function also has a dramatic effect on the "kinetic accessibility" of correctly bound conformations. We explored the parameters that control this component of a successful docking algorithm. We found that docking was more efficient when the energy function was "softened" by eliminating large repulsive nonbonded interactions. However, softening of the potential also diminished the ability of the function to separate correct docking configurations from incorrect ones. Our work on this problem provides guidelines for the development, assessment, and optimization of energy functions and protocols for computational docking studies and should be useful in ongoing ligand design and exploration.

MECHANISMS OF PROTEIN FOLDING

Understanding the means by which a linear amino acid sequence adopts its functional 3-dimensional structure is a key challenge for scientists in many disciplines, from biology to physics. We are using statistical mechanics and computer simulation to elucidate the principles that govern this process. Our explorations of the links between protein topology and the overall mechanism of protein folding are providing new insights that fuel further development of theoretical models and experiments. Our specific focus during the past year was the folding mechanism of simple single-domain proteins with different overall topology. We have computed first-principles free-energy landscapes for the folding of 2 distinct protein topologies, an all-helical protein and a mixed /ß motif.

Figure 1 shows the molecular topologies and the corresponding folding free-energy surfaces from our calculations on fragment B of staphylococcal protein A and segment B1 of streptococcal protein G. These free-energy surfaces present a landscape for folding that connects topology with mechanism. For the helical protein, folding, which is dominated by local interactions, is downhill with concomitant collapse and formation of native tertiary structure as indicated by the "diagonal" nature of the free-energy surface projected onto the R_g and /rhobdy/ reaction coordinates. The folding of protein G involves initial collapse, with only small amounts of native tertiary structure being formed, and then a "search" through compact conformational states for the native structure.

On the basis of these findings and other ongoing work in our laboratory, it appears that this general picture holds. That is, for proteins with more delocalized topologies--ß-sheet structure--folding includes a collapse phase before the acquisition of native structure; more localized structures can fold with concomitant collapse and formation of native structure. Our work in this area is coupled to the theoretical and experimental developments ongoing in other laboratories at TSRI and in the La Jolla area.

PUBLICATIONS

Brooks, C.L. III. Simulations of protein folding and unfolding. Curr. Opin. Struct. Biol. 8:222, 1998.

Brooks, C.L. III, Gruebele, M., Onuchic, J.N., Wolynes, P.G. Frontiers of science 1997: Chemical physics of protein folding. Proc. Natl. Acad. Sci. U.S.A., in press.

Guo, Z., Brooks, C.L. III. Rapid screening of binding affinities: Application of the *-dynamics method to a trypsin-inhibitor system. J. Am. Chem. Soc. 120:1920, 1998.

Guo, Z., Brooks, C.L. III. Thermodynamics of protein folding: A statistical mechanical study of a small ß protein. Biopolymers 42:745, 1997.

Guo, Z., Kong, X., Brooks, C.L. III. Efficient and flexible algorithm for free energy calculations using the *-dynamics approach. J. Phys. Chem. 102:2032, 1998.

Hirst, J.D., Dominy, B., Guo, Z., Vieth, M., Brooks, C.L. III. Conformational and energetic aspects of receptor-ligand recognition. Am. Chem. Soc. Symp. Series, in press.

Reddy, V.S., Giesing, H.A., Morton, R.T., Kumar, A., Post, C.B., Brooks, C.L. III, Johnson, J.E. Energetics of quasi-equivalence: Computational analysis of protein-protein interactions in icosahedral viruses. Biophys. J. 75:546, 1998.

Rychlewski, L., Zhang, B., Godzik, A. Function and fold predictions for Mycoplasma genitalium proteins. Folding Design, in press.

Rychlewski, L., Zhang, B., Godzik, A. Searching for the optimal sequence similarity function. Protein Eng., in press.

Shea, J.-E., Nochomovitz, Y.D., Guo, Z., Brooks, C.L. III. Exploring the space of protein folding Hamiltonians: The balance of forces in a minimalist ß-barrel model. J. Chem. Phys., in press.

Sheinerman, F.B., Brooks, C.L. III. Calculations on folding of segment B1 of streptococcal protein G. J. Mol. Biol. 278:439, 1998.

Sheinerman, F.B., Brooks, C.L. III. A molecular dynamics simulation study of segment B1 of protein G. Proteins 29:193, 1997.

Sheinerman, F.B., Brooks, C.L. III. Molecular picture of folding of a small /alphapub//ß protein. Proc. Natl. Acad. Sci. U.S.A. 95:1562, 1998.

Vieth, M., Hirst, J.D., Brooks, C.L. III. Do active-site conformations of small ligands correspond to low free-energy solution structures? J. Comput. Aided Mol. Des., in press.

Vieth, M., Hirst, J.D., Dominy, B.N., Daigler, H., Brooks, C.L. III. Assessing search strategies for flexible docking. J. Comp. Chem., in press.

Vieth, M., Hirst, J.D., Kolinski, A., Brooks, C.L. III. Assessing energy functions for flexible docking. J. Comp. Chem., in press.

Zhang, B., Jaroszewski, L., Rychlewski, L., Godzik, A. Similarities and differences between non-homologous proteins with similar folds: Evaluation of threading strategies. Folding Design 2:307, 1997.