Speaker: Didier Devaurs, University of Edinburgh

Abstract:

Protein interactions are often associated with changes in a protein's conformation. Studying this structure-function relationship requires gathering information about a protein's conformational space. While experimental techniques have enabled the description of numerous molecular structures, computational methods are required to explore the conformational space of proteins. However, because of the curse of dimensionality, efficient conformational exploration remains a challenge in structural biology. In this talk, I will present two strategies I have used to mitigate the curse of dimensionality when computationally exploring the conformational space of a protein or a molecular complex.
First, I will show how conformational exploration can be guided by low-resolution structural information, such as the experimental data obtained through hydrogen-deuterium exchange (HDX) monitoring. Using experimental data as a bias is a common strategy, but very few approaches use HDX data. Existing methods involve molecular dynamics or atomistic Monte Carlo simulations, which are computationally expensive, especially for large molecular systems. Instead, I followed a coarse-grained conformational sampling approach simplifying the protein model to allow for scalability.
Second, I will show how adopting a purely geometric abstraction allows enhancing the scalability of existing computational methods. I have applied this strategy to the molecular docking of large ligands to protein receptors, and more specifically to the docking of peptide-HLA complexes. I developed a parallelized incremental meta-docking approach, called DINC, that iteratively docks larger and larger overlapping fragments of a ligand in a protein’s binding site. The specificity of my approach is to not break the ligand into biophysically-relevant fragments, but treat it as a purely geometric object and divide it based on what can most benefit the docking of fragments.

Speaker bio:

Didier Devaurs is a cross-disciplinary research fellow (XDF) at the University of Edinburgh. In the past ten years, his focus has been on the computational modelling, simulation and analysis of complex physical systems, both in robotics and structural biology. During his PhD (University of Toulouse, France), he developed novel extensions of sampling-based path planning algorithms by creating the concept of optimal path planning in a cost space. As a post-doctoral researcher (Rice University, USA), he developed computational methods for the conformational modelling and analysis of molecular systems, such as large proteins and protein-peptide complexes. For the past two years (University of Edinburgh), he has been working on several quantitative biomedical research topics, including modelling the intracellular trafficking network of cell-adhesion proteins and statistically analysing time series of white blood cell data. His research currently focuses on addressing quantitative limitations of deep mutational scanning experiments, with the long-term goal of producing clinical interpretations of protein variants in patients using state-of-the-art deep learning methods.