James E. Fitzgerald
In June 2007, I finished my undergraduate degrees at the University of Chicago in physics and mathematics. I am now a graduate student in the Physics Department at Stanford University studying theoretical neuroscience . My undergraduate research focused on computational and theoretical aspects of protein folding jointly with Professors Tobin Sosnick and Karl Freed. During my Junior and Senior years, my research was partially supported by the Arnold and Mabel Beckman Foundation.
- I have been further studying the fundamental motions of proteins by investigating the effects of geometric constraints on protein dynamics. This work has important implications for protein folding pathways and hydrogen exchange mechanisms.
- The energy functions that I have developed in previous work are currently being adapted for optimal performance in a variety of applications. I am focusing mostly on protein structure refinement, but I am also investigating the utility of these functions in the problems of protein structure prediction and protein-protein docking.
- One primary research interest has been the dynamics of proteins. My first project was to investigate fundamental local motions of proteins. Through numerous simulations we find a fundamental anticorrelation between the motions of the phi torsion angle of the ith residue and the psi angle of the (i-1)st residue. This motion is accomplished through the rocking of the rigid peptide group. This motion is independent of solvent model, force field, and is seen both in the presence and absence of long range interactions. This fundamentally local motion is robust enough to describe both equilibrium fluctuations of the backbone and basin transitions. No longer range correlations are necessary in either type of motion to retain the global structure of the peptide. A manuscript of this work is published in the journal Biochemistry.
- In addition to my work on protein dynamics, I have considered the energetic and entropic content of protein sidechains. Using the information learned in this process I have developed new energy functions for use in computational studies of proteins. These energy functions use statistical mechanical concepts to relate statistics compiled from known protein crystal structures to the energy of a protein conformation. The correct expression for this energy requires many-body interactions. However, for computational simplicity we assume that energies are pair-wise additive and depend only on the distances between pairs of atoms. This assumption is obviously incorrect, and in order to diminish the associated errors, we introduce additional conditional dependences to implicitly reproduce these many body effects. Using these considerations, several energy functions are produced which only explicitly include pair-interactions between the heavy backbone atoms and a single side-chain atom (the beta-carbon). However, even in this reduced representation these new energy functions perform better than the best known all-atom energy function based upon their ability to distinguish native from non-native structures (average performance on 306 different decoys sets). I have also built the all-atom versions of these potentials which work even better than the reduced potentials. All potentials are currently in the optimization stage in order to be used for protein structure prediction and structure refinement. This work is published in Protein Science.
- Different simulation methods for studying proteins sometimes give strikingly different results. I am studying popular force fields and solvent models to clearly dilineate these simularities and differences. In particular, I am coauthor on a manuscript (submitted to J. Phys. Chem. B.) in which these comparisons are studied in detail for the beta-amyloid peptide. We find that our implicit solvent model compares reasonably well to experimental measurement for certain physical force fields. There are significant differences between the various physical force fields. We also find that explicit solvent simulations take longer than implicit solvent simulations to reach equilibrium. Given that implicit solvent simulations are also order of magnitides faster than explicit solvent simulations, the ability of our implicit solvent model to reproduce experimental measurements is of practical interest to anyone studying protein dynamics computationally.
- Fitzgerald, J.E., Jha, A.K., Colubri, A., Sosnick, T.R., and Freed, K.F. (2007) Reduced C-Beta Statistical Potentials Can Outperform All-Atom Potentials in Decoy Identification. Protein Science. 16. 2123-2139. 
- Fitzgerald, J.E., Jha, A.K., Sosnick, T.R., and Freed, K.F. (2007) Polypeptide Motions Are Dominated by Peptide Group Oscillations Resulting from Dihedral Angle Correlations Between Nearest Neighbors. Biochemistry. 46. 669-682.