Abstract by Trevor Schmidt
Physics and Astronomy
Dennis Della Corte
Protein Structure Prediction
CASP is a community-wide experiment designed as a blind test for protein structure prediction. The refinement category has a goal of closing the gap between predicted and experimental results. Here some of the predicted conformations submitted are released to refinement groups for further improvement.
The use of averaging in selecting a most probable conformation is prone to introducing several errors. Cartesian averaging of 3D point clouds often results in poor chemical parameters such as bond length. At a larger scale, averaging can produce even greater irrelevancies. One can imagine a domain of a protein that in simulation is shown to adopt two distinct, stable conformations. If the domain is found in each conformation half of the time, then a simple averaging method places the domain in an unobserved middle conformation.
We believe that amongst the plethora of conformational data contained in the MD trajectories, there exist more accurate predictions of the protein’s true structure than we are currently reporting using averaging techniques.