ISBN 9783843938662

978-3-8439-3866-2, Reihe Biophysik

Matthias Ernst
Finding Reaction Coordinates for Protein Folding and Functional Motion

147 Seiten, Dissertation Albert-Ludwigs-Universität Freiburg im Breisgau (2018), Softcover, A5

Zusammenfassung / Abstract

In order to understand how life works at the molecular level, a proper understanding of protein function and dynamics is required. Molecular dynamics simulations allow to study biomolecular dynamics and to assess processes happening within femto- to milliseconds and longer. In order to reduce and analyze the large amount of data generated by these simulations, various methods have been proposed to transform the set of input coordinates to new collective coordinates. These may be used to identify a subset of coordinates which should capture all relevant processes and thus forms a representative multidimensional reaction coordinate. Several issues arise in this process: We may use different choices of input coordinates, like Cartesian atomic positions, dihedral angles or distances between groups of atoms, but it is unknown how this choice affects the overall result. There are different transformations to construct collective coordinates, which in turn may show characteristic properties interpretation of the associated molecular motion may be difficult. Moreover, it is not clear how to choose the subset of collective coordinates and which descriptors allow to select meaningful reaction coordinates in the first place.

In this work, we focus on principal component analysis as method for the generation of collective coordinates. We present a way to employ inter-residue distances as input coordinates in a systematic way and compare its performance to the established methods, where Cartesian coordinates or dihedral angles are used. Adopting three model systems which show various types of dynamics we find that the choice of suitable input coordinates crucially depends on the major type of dynamics: For the folding system villin headpiece (HP35), backbone dihedral angles are suited best. T4 lysozyme on the other hand shows a large-scale domain motion, which is not captured by backbone dihedral angles, but described well by inter-residue distances. Thus, we argue that it is beneficial to evaluate simple parameters prior to complicated analysis to determine the dominant motions and then choose coordinates which resolve it best. Finally, we use simulations where we pull parts of the molecule together or apart as a “molecular tweezer” to show that the slow large-scale domain motion of T4 lysozyme causally depends on fast motion of a single side chain, which acts as a lock. We thus present a promising approach to investigate hierarchical dynamics of biomolecules.