Structure evaluation is a necessary component of both de novo structure prediction and modeling by homology. Protein structures are currently being evaluated by several methods: in vacuo energy minimization, all atom free energy calculations, empirical solvation functions, pair potentials, and environmental preferences. Evaluations of protein structures have been successful in some cases, but failed in others. To improve on these evaluations, I am using additional information provided by multiple sequence alignments. I use multiple alignments formed by searching a sequence database with the sequence of the structure to be evaluated (structure sequence). I retain sequences which should fold to the same overall structure (related sequences). I align all the sequences with the structure sequence, so each amino acid in the structure sequence (structure amino acid) matches several amino acids (related amino acids) from the aligned related sequences. Each of the protein structures that correspond to one of the related sequences is slightly different. However, their level of similarity is such that at each position in a multiple alignment, most of the related amino acids experience a similar structural environment. Additionally, the level of variability in related amino acids at aposition in a multiple alignment can be used to predict surface exposure. Thus variation in amino acids at a position in the multiple alignment along with the structural amino acid can be used to evaluate a position in a model structure. I reclassify each structure amino acid into four groups based on the related amino acids: variable polar and variable nonpolar, variable polar and invariant nonpolar, invariant polar and variable nonpolar, invariant polar and invariant nonpolar. The propensity of each reclassified amino acid to be in several different environments is calculated with a simple statistical procedure. I sum the propensities of each position in the model structure to evaluate the structure. This approach has been quite successful, and a paper should be forthcoming. I have also assisted in building a model structure for the factor eight protein. Mutations in this protein are associated with hemophilia. This model has led to a project developing a mutant form of factor eight which may have a longer blood circulation time, easing the lives of hemophiliacs on replacement factor eight. LP I also evaluated the structure predictions at the structure prediction Asilomar. This study will serve as a benchmark for the ab-initio structure prediction field. Each of these projects extensively used MidasPlus molecular visualization software, both for research, and development of publication quality figures. In addition, I extensively use the sequence analysis package Eugene.
Showing the most recent 10 out of 508 publications
