- Email: [email protected]
A graph theory approach to predicting unfolding pathways of proteins
retical models, we can show how evolutionary considerations can explain many of the observed properties of proteins such as the way proteins fold, the distribution of observed protein structures, the marginal stability and flexibility of proteins, and how the evolutionary robustness of protein structures co-exists with sequence plasticity.
Molecular Dynamics Simulations on Protein Folding and Protein Structure Prediction Peter Kol/man, Department of Pharmaceutical Chemistl T, University of California, San Francisco, CA 94143 USA
We will review our recent work on molecular dynamics simulations using all atom force fields with solvent and counterions on protein folding. Simulations on the first phase of protein folding have been carried into the microsecond regime for the protein villin and into the hundreds of nanosecond range for the protein B B A I . Secondly, we describe the combined use of molecular dynamics and free energy calculations using the Molecular MechanicsPoisson Bolzmann Surface Area approach, with the goal of discriminating decoys from correct structures in protein folding.
A Graph Theory Approach to Predicting Unfolding Pathways of Proteins Leslie A. Kuhn, 1.2Brandon Hespenheide, i.: Andrew J. Radel; 1.2 and M.E Thorpe, 2 Departments of Biochemistry t and Physics and Astronomy, 2 Michigan State University East Lansing, MI 48824 USA. http://www.bch.msu.edu/labs/kuhn and http ://u,ww.pa.nlsu.edu/-thorpe
Protein folding is driven by hydrophobicity, but once compact states are achieved, hydrogen bonds tend to lock in the structure. This is particularly apparent in the regular networks of hydrogen bonds in ct-helices and B-sheets, with cross-linking side-chain hydrogen bonds and salt bridges also contributing. Here, we investigate a new hypothesis: that significant information about the folding pathway is preserved in the density and strength of hydrogen bonds in the folded protein, as observed in its three-dimensional structure. This is based on the idea that the strongest interactions are likely to occur first during folding as well as being the hardest to break afterwards. Therefore, strong hydrogen bonds, or a high density of weaker hydrogen bonds, build upon favorable van der Waals and hydrophobic interactions in a region to lock its structure, creating a substructure that tends to persist along the folding pathway. While this is difficult to test by folding simulations, we can simulate the thermal unfolding of the hydrogen-bond network by breaking these bonds one by one, from weakest to strongest, as though the protein were being heated gradually. At each step, the rigid (structurally stable) and flexible (unstable) regions in the protein are determined by graph-theoretic analysis of its covalent and hydrogen-bond network using FIRST (see abstracts by Thorpe, Hespenheide, Rader, Jacobs, Krishnamurthy, and Lei). This approach allows us to test whether the order in which hydrogen-bonded substructures become unstable corresponds to
the order in which they unfold, according to hydrogenexchange NMR and limited proteolysis experiments.
Comparing Protein Structures: A GaussianBased Approach to the Three-Dimensional Structural Similarity of Proteins Gerald M. Maggiore, Douglas C. Rohrer, and Jordi Mestres, I Computer-Aided Drug Discovery, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, M149007-4940, USA, ~Department of Molecular Design & lnformatics, N. V. Organon, P.O. Box 20, 5340 B.H Oss, The Netherlands. [email protected], [email protected], [email protected]
A new method for comparing three-dimensional protein structures based on an optimal alignment of their "steric fields" is described. The method is based upon the use of spherical Gaussian functions located on individual atoms. This representation generates a flexible description of the underlying fold geometry of proteins that can be adjusted by changing the width of the Gaussian functions. Reducing the width sharpens the representation and leads to a more atom-like description; increasing the width creates a fuzzier representation that preserves the general shape features of the chain fold but with a consequent loss in atomic resolution. The width used in the present work, which is based upon the features of individual atoms, provides a representation that is seen to be quite robust with respect to the variety of geometric features typically encountered in the alignment process. In addition, a post-alignment analysis is performed that generates sequence alignments from the corresponding structure alignments. An example, based on five mammalian and fungal matrix metalloproteinases (human fibroblast collagenase, neutrophil collagenase, stromelysin, astacin, and adamalysin), is given that illus-
trates a number of features of the Gaussian-based approach.
Protein Structure Prediction John Moult, CenterJbr Advanced Research in Biotechnology, University of Mal3'land Biotechnology hlstitute, 9600 Gndelsky Dl:, Rockville, MD 20850 USA
For over 30 years, the task of predicting protein structure from amino acid sequence has been a grand challenge problem in computational biology. Although we are still far from a solution that can compete with experiment, much progress has been made, and it is now possible to produce some sort of model for a significant fraction of known sequences. How accurate are these models, what are the real obstacles to making them better, and what do they teach us about the nature of protein structures? CASP, a community wide system of experiments to measure the performance of prediction methods, has provided a large body of data with which to address these questions. I will attempt to use these data to identify the limits of present prediction techniques, and will discuss those limits in terms of the underlying physical and organizational properties of protein molecules.
J. Mol. Graphics Mod., 2000, Vol. 18, August-October
553
Molecular Dynamics Simulations on Protein Folding and Protein Structure Prediction Peter Kol/man, Department of Pharmaceutical Chemistl T, University of California, San Francisco, CA 94143 USA
We will review our recent work on molecular dynamics simulations using all atom force fields with solvent and counterions on protein folding. Simulations on the first phase of protein folding have been carried into the microsecond regime for the protein villin and into the hundreds of nanosecond range for the protein B B A I . Secondly, we describe the combined use of molecular dynamics and free energy calculations using the Molecular MechanicsPoisson Bolzmann Surface Area approach, with the goal of discriminating decoys from correct structures in protein folding.
A Graph Theory Approach to Predicting Unfolding Pathways of Proteins Leslie A. Kuhn, 1.2Brandon Hespenheide, i.: Andrew J. Radel; 1.2 and M.E Thorpe, 2 Departments of Biochemistry t and Physics and Astronomy, 2 Michigan State University East Lansing, MI 48824 USA. http://www.bch.msu.edu/labs/kuhn and http ://u,ww.pa.nlsu.edu/-thorpe
Protein folding is driven by hydrophobicity, but once compact states are achieved, hydrogen bonds tend to lock in the structure. This is particularly apparent in the regular networks of hydrogen bonds in ct-helices and B-sheets, with cross-linking side-chain hydrogen bonds and salt bridges also contributing. Here, we investigate a new hypothesis: that significant information about the folding pathway is preserved in the density and strength of hydrogen bonds in the folded protein, as observed in its three-dimensional structure. This is based on the idea that the strongest interactions are likely to occur first during folding as well as being the hardest to break afterwards. Therefore, strong hydrogen bonds, or a high density of weaker hydrogen bonds, build upon favorable van der Waals and hydrophobic interactions in a region to lock its structure, creating a substructure that tends to persist along the folding pathway. While this is difficult to test by folding simulations, we can simulate the thermal unfolding of the hydrogen-bond network by breaking these bonds one by one, from weakest to strongest, as though the protein were being heated gradually. At each step, the rigid (structurally stable) and flexible (unstable) regions in the protein are determined by graph-theoretic analysis of its covalent and hydrogen-bond network using FIRST (see abstracts by Thorpe, Hespenheide, Rader, Jacobs, Krishnamurthy, and Lei). This approach allows us to test whether the order in which hydrogen-bonded substructures become unstable corresponds to
the order in which they unfold, according to hydrogenexchange NMR and limited proteolysis experiments.
Comparing Protein Structures: A GaussianBased Approach to the Three-Dimensional Structural Similarity of Proteins Gerald M. Maggiore, Douglas C. Rohrer, and Jordi Mestres, I Computer-Aided Drug Discovery, Pharmacia Corporation, 301 Henrietta Street, Kalamazoo, M149007-4940, USA, ~Department of Molecular Design & lnformatics, N. V. Organon, P.O. Box 20, 5340 B.H Oss, The Netherlands. [email protected], [email protected], [email protected]
A new method for comparing three-dimensional protein structures based on an optimal alignment of their "steric fields" is described. The method is based upon the use of spherical Gaussian functions located on individual atoms. This representation generates a flexible description of the underlying fold geometry of proteins that can be adjusted by changing the width of the Gaussian functions. Reducing the width sharpens the representation and leads to a more atom-like description; increasing the width creates a fuzzier representation that preserves the general shape features of the chain fold but with a consequent loss in atomic resolution. The width used in the present work, which is based upon the features of individual atoms, provides a representation that is seen to be quite robust with respect to the variety of geometric features typically encountered in the alignment process. In addition, a post-alignment analysis is performed that generates sequence alignments from the corresponding structure alignments. An example, based on five mammalian and fungal matrix metalloproteinases (human fibroblast collagenase, neutrophil collagenase, stromelysin, astacin, and adamalysin), is given that illus-
trates a number of features of the Gaussian-based approach.
Protein Structure Prediction John Moult, CenterJbr Advanced Research in Biotechnology, University of Mal3'land Biotechnology hlstitute, 9600 Gndelsky Dl:, Rockville, MD 20850 USA
For over 30 years, the task of predicting protein structure from amino acid sequence has been a grand challenge problem in computational biology. Although we are still far from a solution that can compete with experiment, much progress has been made, and it is now possible to produce some sort of model for a significant fraction of known sequences. How accurate are these models, what are the real obstacles to making them better, and what do they teach us about the nature of protein structures? CASP, a community wide system of experiments to measure the performance of prediction methods, has provided a large body of data with which to address these questions. I will attempt to use these data to identify the limits of present prediction techniques, and will discuss those limits in terms of the underlying physical and organizational properties of protein molecules.
J. Mol. Graphics Mod., 2000, Vol. 18, August-October
553