Berlin 2018 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

BP: Fachverband Biologische Physik

BP 15: Postersession III

BP 15.101: Poster

Tuesday, March 13, 2018, 14:00–16:00, Poster B

Comparison of Coevolutionary Protein Structure Prediction Methods — •Lars Franke¹ and Alexander Schug² — ¹Karlsruhe Institute of Technology (KIT) — ²Jülich Supercomputing Centre (JSC)

Analyzing the structure of proteins is the key to understanding the biological mechanisms they are involved in. While the experimental structure determination of proteins is expensive, the corresponding genetic sequences are easy to obtain. Correlated mutations in the sequences within a protein family provide information about the protein's function and therefore its structure. These correlations can be translated into residue-residue interactions using Direct Coupling Analysis (DCA). We generate contact predictions for eight proteins with known structure with three different methods---mean field DCA, Pseudo Likelihood Maximization and Deep Neural Networks---and compare the results. This study can provide orientation in the world of protein structure prediction tools.