Grid'5000 user report for Alexandru Tantar

Jump to: navigation, search

User information

Alexandru Tantar (users, user, account-manager, luxembourg, ml-users user)
More user information in the user management interface.


  • DOCK - Conformation Sampling and Docking on Grids, Docking@GRIDS (Other) [achieved]
    Description: Molecular modeling­ and notably the conformational sampling and docking procedures­ are in principle able to provide help for understanding the interaction mechanisms between (macro)molecules involved in physiological processes. The processes to simulate are of a combinatorial complexity (molecule size, number of degrees of freedom) that represents an important challenge for the currently available computing power, hence the three imperative research directions in present day's molecular modelling: (1) the search for mathematical models of maximum simplicity that nevertheless provide a relevant description of molecular behaviour, (2) the development of powerful distributed optimization algorithms (genetic algorithms, local search, hybrid algorithms) for sampling the molecular energy surface for stable, populated conformations, and (3) deploying those intrinsic distributed algorithms on computational Grids. This approach will be generalized to include intermolecular degrees of freedom, turning it into a flexible docking procedure.

    The computational prediction of energetically stable molecular conformations is at the core of computer-aided molecular property and biological activity predictions. Such predictions are of obvious economic interests, as their success would significantly alleviate the need to randomly synthesise and test molecules in order to discover the compound with desired properties. The problem of optimizing the energy function of the geometry is NP-complete with tens to thousands of degrees of freedom and would, per se, justify the use of distributed computing. A more fundamental problem than the sampling of the potential energy surface is its accurate definition – classical “force fields” being either too specialized (for proteins) or inaccurate. Therefore, classical docking algorithms typically rely on a “docking” function to position the ligand in the protein binding site, and a different empirical “scoring” function to estimate binding affinity on hand of the previously obtained poses. This is incoherent, but necessary unless fully flexible docking is performed. While it is clear that docking methodology needs further improvement, most of it is expected to come from a wide-scale development and employment of flexible docking procedures, accounting for conformational flexibility of both site and ligand. However, this is not feasible without appealing to the development of docking procedures specifically adapted to the GRID which is certainly an important research direction for the coming years.

    The first challenge of this project is to develop novel multimodal GRID optimisation algorithms that are well suited for distributed computing and specifically adapted to conformational sampling and docking, based on a genetic algorithm hybridized with other optimisation heuristics. The second challenge of the current project is to seek for a generally applicable force field that allows docking free energies to be estimated on hand of the energy levels of the sampled conformers of free and bound states (in other words, to reduce docking to a conformational sampling problem of two molecules). This implies a sampling and docking tool on the GRID, powerful enough to allow multiple simulations of various learning set molecules (conformational sampling of structured peptides) and complexes (docking) with different force field parameter sets, in quest of a force field setup returning accurate free energies for all systems under study. This would be, to our knowledge, a first attempt to parameterise a force field according to its propensity to return correct ensemble properties of a simulated system (rather than correct local energy barriers in small molecules).
    Results: A number of results may be found in the mentioned articles.
    More information here
  • Sparse Antenna Array Optimization - CEA CEST, ALEA, INRIA Bordeaux - Sud-Ouest (Other) [achieved]
    Description: French Atomic Energy Commission (CEA), project coordinated by Prof. Del Moral (Advanced Learning Evolutionary Algorithms, INRIA Bordeaux – Sud-Ouest / IMB) and Pierre Minvielle (CEA CESTA, Le Barp). External collaboration as a member of the Advanced Learning Evolutionary Algorithms (ALEA) Team, offering support on the design and analysis of different evolutionary approaches. A significant improvement of the formerly obtained results was attained, superseding the cross-entropy based approaches, previously addressed in the project. Sparse antenna arrays stand as a high interest topic in the electromagnetic measures domain, communications, etc. From a formal point of view, the optimization of a sparse antenna array, with respect to various constraints, can be modelled over a set of continuous functions describing directivity, lobes, etc. Nonetheless, as a result of the non-convex and highly multi-modal nature of the functions to be optimized, classical algorithms are generally ineffective. Extending previous approaches, a Kullback-Leibler cross-entropy based stochastic paradigm is considered for the study, the algorithm being conducted by performing adaptive changes of the probability density functions in order to explore the search space.
    Results: A significant improvement of the formerly obtained results was attained, superseding the initial approaches used by CEA CESTA. The results are subject to publication as part of an IEEE Transactions journal paper.
    More information here



    Success stories and benefits from Grid'5000

    last update: 2011-06-22 11:20:00

    Personal tools

    Public Portal
    Users Portal
    Admin portal
    Wiki special pages