Graph representation of protein 3D conformational ensemble, application to docking

Type de poste
Durée du poste
Contrat renouvelable
Contrat non renouvelable
Date de prise de fonction
Date de fin de validité de l'annonce

<p><span style="color: rgb(34, 34, 34); font-family: arial, sans-serif; font-size: 13px; background-color: rgb(255, 255, 255);">Campus Scientifique, 615 Rue du Jardin-Botanique, 54506 Vandœuvre-lès-Nancy</span></p>

Isaure Chauvot de Beauchene
Dave Ritchie
Email du/des contacts

<div><div><strong>&nbsp;Job environnement</strong></div><div>&nbsp;</div><div>The research of the CAPSID team ( combines aspects of various disciplines : biophysics, structural biology, and algorithmic related to 3D coordinates, graphs, clustering and machine learning. We have recently developed an original methodology to build atomic representations of single-stranded nucleic acids bound to a protein, using a fragment-based approach. The method is uniquely capable of providing accurate structural model from only protein structure and RNA sequence. The goal of this post-doc project is to generalise this method towards proteins of approximately known structure from homology. To account for its flexibility, we will represent the protein as a graph of possible local conformations at different scales. This development will be based on molecular docking tools existing and / or developed in the CAPSID team.</div><div>&nbsp;</div><div>&nbsp;</div><div><strong>Context</strong></div><div>&nbsp;</div><div>Many protein structures have been determined experimentally and similar sequences tend to adopt similar 3D conformations, especially in conserved domains. While enabling the use of homology modelling, this still leaves room for a lot of uncertainty or variability in the protein conformations, at different scales: Several domains in a protein are linked by flexible loops and can be differently oriented; They contain themselves flexible loops with multiple possible backbone conformations ; All residues can adopt different side-chain orientations.</div><div>&nbsp;</div><div>In classical docking (i.e. modeling of a molecular assembly from the structure of its constituents), an ensemble of protein conformations can be sampled in order to increase the chances get close enough to the protein bound form, But this requires a combinatorial sampling of each flexibility scale and each protein region, and can therefore explore only a very limited part of the conformational space. In our fragment-based approach, the conformations of each binding site of the receptor can be sampled independently, and their compatibility assessed during the linear fragment assembly process.</div><div>&nbsp;</div><div>&nbsp;</div><div><strong>Project description</strong></div><div>&nbsp;</div><div>We will exploit this linearity property on an archetypal domain: the RNA-recognition motif (RRM). While the RRM is estimated to be present in 1-2 % of the human genome, its RNA-recognition code has not been fully cracked yet. Using the wealth of existing knowledge on RRM atomic structures, local 3D connectivity patterns between RRM regions and between RRM and ssRNA fragments will be identified. These patterns will then be integrated into the construction of a global connectivity graph of ssRNA fragments (using the approach previously developed in our group). This graph will provide an atomic-level description of the RRM-ssRNA interaction.</div><div>&nbsp;</div><div>We will first use the approximation that loops are independent of each other, and that only spatially close rotamers are interdependent. Second, we will add as a constraints the knowledge of one or two key amino-acid - nucleotide contacts. Rotamers dependencies and key contacts will be identified by (co)-evolution in protein sequence alignments. In addition, new sequence-based predictors will be developed by machine learning applied on improved multiple sequence alignments and HMM approaches, in collaboration with the lab of Dr. Wim Vranken at the Vrije Universiteit Brussels, Belgium.</div><div>&nbsp;</div><div>If the precision obtained previously is not satisfying enough, we will use (simulated) low resolution data on the geometry of the complex, that could be obtained experimentally (SAXS, NMR), as additional constraints.</div><div>&nbsp;</div><div>&nbsp;</div><div><strong>Required qualifications</strong></div><div>&nbsp;</div><div>The project is interdisciplinary. Candidates must have (or be about to get) a PhD degree in any of the relevant disciplines: (bio-)physics, bio-informatics, computer science or structural biology. Strong programming skills (preferentially Python and/or C++) are required. Skills in discrete mathematics, statistics and/or knowledge of molecular structures are very desirable.</div><div>&nbsp;</div><div>Candidates must be fluent either in French or in English</div><div>&nbsp;</div><div>&nbsp;</div><div><strong>Application</strong></div><div>&nbsp;</div><div>To fill up on</div><div>Deadline : 8th June 2018</div><div>&nbsp;</div></div>
<style type="text/css">
Laboratoire: LORIA (CNRS - INRIA - UL)