Detailed Modeling in Structural Genomics and Single Molecule Dynamics
Computer modeling becomes more and more important in structural
genomics and protein dynamics study. We have developed computer programs
for atomic detailed modeling in protein structure and function prediction.
In protein structure prediction, atomic detailed statistical potential
has been shown to perform better in near-native structure selection
and in structure refinement. In a test set of 65 proteins, the current
method can generate near-native structures for 75% of the cases. The
statistical potential guided refinement procedure can improve the protein
structure for half of the test cases, while keeping the other half unchanged.
In structure-based function prediction, we have developed multiresolution
statistical potentials and a complex threading algorithm in protein-protein
interaction study. In both structure and function prediction, our atomic
modeling programs are sufficiently fast for genome scale applications.
Finally, an atomic detailed steered molecular dynamics simulation is
used in studying protein's mechanical property. Combined with the results
from single molecule experiments, we have investigated protein's dynamics-function
relationship. The discovery of a hydrogen bond controlled unfolding
mechanism has been confirmed by protein engineering and atomic force