PhD Position on Interactive Simulations

(A PDF version is available by clicking here)

The interactive molecular simulations approach offers user to visualize in real time the result of a running simulation and apply different force constraints over few atoms. These constraints are taken into account during the simulation, allowing molecule of interest manipulation in order to model it, but also to address relevant challenges in the field of drug design.

However, despite the huge amount of computing performances available, the method used as reference, all-atoms molecular simulation, shows some limitations, foremost to deal with large crystallised structures, but also when applied to specific events time scale a researcher would supposes to observe (sometimes at the scale of one second, whereas we reach only few microseconds in the most ambitious simulations). Then, going through more efficient calculations, adapted to a coarser scale but still relevant on a biophysical point of view, appears to be essential. It would allow, on the first hand, to limit the hypothesis that must be tested by more expensive simulation, and, on the other hand, thanks to its interactive ability, to address modelisation and in silico building issues when dealing with large and complex structures.


In the LIMSI and the LBT, we are conjointly developing a well-adapted approach, able to address very complex molecular structures at larger time scales, compatible with high-performance constraints imposed by the interaction. The aim of this thesis will be then to refine this multi-scale approach based on a modelisation of biomolecules by spring networks, known to accurately render molecular mechanical properties. This apporach was inspired by the normal modes analysis and can be apply whatever the grain size we use (atom, coarse grain, residue, ...) and the biomolecule type modelled (protein, RNA, DNA, ...). This modelisation is supplemented by the integration of non-covalent interactions (steric and electrostatic) also applicable to different modelisation scales. Finally, the approach also integrates at a thinner scale a modelisation of proteins as articulated rigid bodies, technic inspired by robotic and high-performance physic engines used in video games. Moreover, they allow the global and local parameterization of a protein flexibility thanks to structural knowledge obtained from experiments or homology methods.

This approach, concretised in Biospring software (parallelised code with OpenMP and OpenCL, for which few processes have been partially parallelised on CPU and GPU with HMPP during portage project GENCI in 2010 – PPLC3D) already proved itself as an efficient approach for interactive modelisation in theoretical biochemistry [1] [2] [3], but needs a better transition to specific scales via more advanced optimisation technics as well as an hybrid portage on multicore (MPI/OpenMPI) and manycore (GPU-OpenCL). This transition to another scale will especially allow considering very large biological complexes like viral capsids, posing new challenges in term of performance. Furthermore, the last on going developments begin to integrate new experimental data like molecular envelops (SAX and CryoEM) bringing up a new reflexion on the most adapted data structures for efficient, performing and portable computing.

The partners LIMSI (Nicolas Férey) and LBT (Marc Baaden) will bring their theoretical and technical skills for the development of the biophysical code. The LBT will also propose a large range of biological applications to define priorities for the developments to be performed (membrane fusion phenomenon, neurotransmission, biological cues transmission, virus infection, ...). They will also valid the relevance of the approach to address the different issues previously evoked. The expertise of the Maison de la Simulation for algorithmic optimization and parallelization will be indispensable for the code transition to this new scale, which will be the main concern of this thesis.

Références :

[1] O. Delalande, N. Ferey, B. Laurent, M. Gueroult, B. Hartmann, and M. Baaden. Multi-resolution approach for interactively locating functionally linked ion binding sites by steering small molecules into electrostatic potential maps using a haptic device. In Proceedings of Pacific Symposium on Biocomputing (PSB'10), volume 15, p. 205-215, 2010.
[2] A. Saladin, C. Amourda, P. Poulain, N. Ferey, M. Baaden, M. Zacharias, O. Delalande, and C. Prevost. Modeling the early stage of dna sequence recognition within reca nucleoprotein filaments. Nucleic Acid Research, 38(19):6313-6323, 2010,,

[3] A. Tek, M. Chavent, M. Baaden, O. Delalande, P. Bourdot and N. Ferey (2012). Advances in Human-Protein Interaction - Interactive and Immersive Molecular Simulations, Protein-Protein Interactions - Computational and Experimental Tools, Weibo Cai and Hao Hong (Ed.), ISBN: 978-953-51-0397-4, InTech.
[4] Rader AJ, Chennubhotla C, Yang L-W, Bahar I, in "Normal Mode Analysis. Theory and Applications to Biological and Chemical Systems, The Gaussian Network Model: Theory and Applications." Eds Cui Q, Bahar I, Math & Comp. Biology Series, Chapman & Hall CRC Press, Taylor & Francis Group, p. 41-64, 2006.

Projets connexes :

  1. [5]  Projet ANR Exaviz (ANR-11-MONU-003),

  2. [6]  EquipEx Digiscope,

  3. [7] EquipEx CACSICE,


Joomla templates by a4joomla