SNIC
SUPR
SNIC SUPR
A computational approach to protein dynamics
Dnr:

SNIC 2018/3-26

Type:

SNAC Medium

Principal Investigator:

Pär Söderhjelm

Affiliation:

Lunds universitet

Start Date:

2018-05-02

End Date:

2019-06-01

Primary Classification:

10407: Theoretical Chemistry

Secondary Classification:

10602: Biochemistry and Molecular Biology

Allocation

Abstract

Dynamics is fundamental for the function of proteins. Experimentally, protein dynamics can be studied in various types of nuclear magnetic resonance (NMR) experiments, but they give limited information about the detailed nature of the fluctuations. Molecular simulations, on the other hand, provides atomic resolution, but the possible length of the simulations is typically around a microsecond, and thus slower fluctuations, which are the key to most biological processes, will never be observed during the simulations. In this project, we will use enhanced sampling methods to make slow fluctuations appear even in simulations of limited length. These methods artificially enhance the fluctuations along all or a few predefined degrees of freedom (usually denoted collective variables). Specifically, we will work within the metadynamics framework and develop a reliable and partly automatic protocol for selecting the best suitable collective variables to address various kinds of protein dynamics. As part of this method development, we are applying the methods to well-defined problems, such as the nature of local unfolding events in gPW, conformational gating in aspartic proteases, and difficult cases of protein-ligand binding. For example, our extensive calculations in the previous allocation period has led to a better understanding of the flap dynamics in the Plasmodium protease (manuscript in preparation). Due to their simple functional form, standard force fields have a limited accuracy, especially when applied to structures outside the parametrization domain (typically crystal structures), such as those visited in enhanced-sampling simulations. Therefore, it is important to explore new types of force fields. A particularly promising development is the Amoeba force field, which is based on multipole electrostatics and explicit polarization. Recent algorithmic developments have made it tractable to study protein dynamics with Amoeba on GPUs. As a part of the current project, we will run simulations with Amoeba and compare the results with standard force fields, thereby contributing to the much needed knowledge of how well improved potentials perform in practise.