Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 21: Poster IIIb
BP 21.13: Poster
Wednesday, March 20, 2024, 11:00–14:30, Poster C
Application of similarity measures to MD simulation data — •Fabian Schuhmann1, Leonie Ryvkin2, James D. McLaren3, Luca Gerhards3, and Ilia A. Solov’yov3 — 1University of Copenhagen, Copenhagen, Denmark — 2Technische Universiteit Eindhoven, Eindhoven, Netherlands — 3Carl von Ossietzky Universität Oldenburg, Oldenburg, Germany
Biological processes involve movements across all measurable scales, which must be analyzed and understood to derive Nature’s reasoning and understand the studied object’s potential function. Especially in molecular dynamics simulations, considerable resources are allocated to get a picture of the motion of a protein.
While one can easily compare a protein structure to a reference employing tools like the Root Mean Square Deviation, methods need to become more involved to compare two whole trajectories. In a stop-motion movie, how does one spot the difference among thousands of atoms, all wiggling and moving?
We have gathered eight different similarity measures in an easy-to-use Python package called SiMBols. SiMBols includes the Hausdorff distance, the (weak) Fréchet distance, dynamic time warping, Longest Common Subsequence, a difference distance matrix approach, Wasserstein distance, and Kullback-Leibler divergence and combines them in a unified way.
Employing a case study, we will use the measures. We will find that the different similarity measures differ in their computation time and the research question they might answer.
Keywords: Molecular Dynamics; Similarity Measure; Data Analysis; Protein Dynamics; Conformational Changes