Petascale integrative approaches to protein structure prediction

Ken Dill, SUNY at Stony Brook

Usage Details

Ken Dill, Alberto Perez, Lane Votapka, James Robertson, Emiliano Brini, Cong Liu, Roy Nassar, Bhanita Sharma, Antonio Bauza Riera

This project addresses a central challenge of biology: the prediction of protein structure from sequence. Accurate determination of protein structure is essential to understanding biological mechanism and designing new drugs. The purpose of this work is to use the petascale capabilities of Blue Waters to overcome limitations of current experimental and computational tools of structure determination. The project will do this by running a physics-based simulation method, called MELD (Modeling Employing Limited Data) that rationally incorporates sparse, noisy, and ambiguous experimental or heuristic information. Currently around 15% of protein targets cannot be tackled by known methods. The transformative goal of this work is to use MELD, running on Blue Waters, to break substantial barriers in size and complexity for computational structure prediction.

This work has three major goals: (1) use MELD to provide structures for the known, small proteins were bioinformatics methods fail and where databases are sparse (e.g. membrane proteins), (2) provide kinetic and pathway information for protein folding by seeding Markov State Models with MELD intermediate states, and (3) use accurate relative free energies of folding upon binding for more accurate peptide design. The project will use the the petascale capabilities of Blue Waters to push the size boundaries of full atomistic physics-based predictions. Both the software and the results obtained from this project will be freely accessible to the larger scientific community.