Albany 2013: Book of Abstracts

category image Albany 2013
Conversation 18
June 11-15 2013
©Adenine Press (2012)

Protein Structure Prediction with limited data

Proteins need to interact with other molecules in order to carry out their biological role. Knowing the protein structure is crucial to study these interactions and it can for example lead to drug design. However, the cost of determining the structure of a protein with current experimental techniques is very high -- both in time and money. In the absence of experimental structures, current computational tools are not always able to correctly predict the native fold.

We are using a physics based computational framework to determine the structure of proteins. The pipeline is designed to handle sparse data coming from evolution, bioinformatics or experiments (solid state NMR, cross linking, …). The data is transformed into a set of restraints used in our physical simulations. However, we require that only a subset of the input information is satisfied. This is done to account for uncertainties in the input data. The framework uses a Hamiltonian-Temperature replica exchange formalism that allows the system to choose what data is compatible with the physics of the system. I will show some results on how this methodology can help us in both protein structure refinement and protein structure prediction.

Alberto Perez
Justin MacCallum
Ken Dill

Laufer Center for Physical and Quantitative Biology
Stony Brook University
Stony Brook NY 11794