Accurate prediction of the three-dimensional structure of a protein based on sequence information is an important step on the way to understanding the function and interactions of a protein. Using machine learning techniques combined with physical force field, we are developing new algorithms to improve the structure prediction of protein sequences that cannot be modeled using homology modeling.
Protein folding and interactions
Understanding the way in which a protein folds reliably and rapidly from an extended chain into its well defined three dimensional structure remains a challenge. We study the mechanisms of folding and protein-protein interactions for small single-domain proteins using molecular modeling with Monte Carlo algorithms.
Biological simulations are still underrepresented in high-performance computing. We are developing algorithms that can take advantage of thousands of cores to take advantage of todays and future high-performance computing platforms. We also develop interfaces that make these and other available programs readily usable for biologists.