Unraveling protein functional motions using Self-Organizing Maps

Speaker: Guillaume Bouvier, Institut Pasteur, Dept. of Structural Biology and Chemistry
Host: Thomas Hamelryck, Section for Computational and RNA Biology

Abstract
Sampling the conformational space of biological macromolecules (e.g. proteins) generates large data sets with considerable complexity. Data-mining techniques, such as clustering, can extract meaningful information. Among them, the self-organizing maps (SOMs) algorithm has shown great promise, in particular since its computation time rises only linearly with the size of the data set. The algorithm has been applied to 3 different fields in structural biology: 1. Determining 3D structures from sparse and heterogeneous data: This point is illustrated by the modeling strategy used to predict an assembly model for the Type 2 Secretion System (T2SS) found in a wide variety of pathogenic and non-pathogenic Gram-negative bacteria. The coupling of electron microscopy data and SOM analysis of a Monte-Carlo based molecular sampling allowed to propose a dynamic model for the T2SS assembly. This model has been validated by experimental data. 2. Analyzing 3D structures to access experimentally unreachable structures and propose transition path between conformational states: The SOM analysis of several molecular dynamics simulations of VanA, an enzyme implied in antibiotic (vancomycin) resistance, has linked the conformational change of the protein with the reaction catalyzed. 3. Exploiting 3D structures to develop new drugs: virtual screening and drug design: The SOM based approach can also be used to cluster docking poses of small molecules to a protein. This strategy has been shown to enhance the sensibility of virtual screening approaches. Our future goal is to use the SOM mapping to guide the sampling to new region of the conformational space. The so called adaptive sampling, will allow to decipher the conformational landscape of larger biological macromolecules.