Berlin 2018 – scientific programme

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

BP: Fachverband Biologische Physik

BP 15: Postersession III

BP 15.56: Poster

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

MaxEnt-Stress Graph Drawing in Protein Structure Determination — •Oskar Taubert¹ and Alexander Schug² — ¹Karlsruhe Institute of Technology, Karlsruhe, Germany — ²Jülich Supercomputing Centre, Jülich, Germany

Proteins perform a wide range of functions in living systems, such as transport, catalysis, or signaling. A protein’s function and structure are closely related. There are different experimental as well as computational methods for resolving protein structures. NMR in particular supplies a list of atom pairs with associated distance intervals and confidence scores. This information has to be translated to 3D atomic coordinates, solving a distance geometry problem.

Since graphs are suited to model this type of pairwise relationships as edges and vertices, we use MaxEnt-Stress graph drawing as an efficient solution to map a list of atomic distance constraints onto a 3D-structural model (cf. Wegner et al., ESA 2017). More specifically, we take input errors and distance intervals into account. Input consists of the amino acid sequence, secondary structure, and long range contact information, as it is provided by experiments or co-evolutionary analysis. To test our algorithm, we conduct simulations on input data generated from known structures of biomolecules of different types. We find the reference structure is reproduced with high fidelity, even from noisy data, when supplying roughly three times the number of heavy atoms as graph edges.