Skip to Main content Skip to Navigation
Habilitation à diriger des recherches

Algorithmes Haute-Performance pour la Reconnaissance de Formes Moléculaires

David Ritchie 1
1 ORPAILLEUR - Knowledge representation, reasonning
INRIA Lorraine, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
Abstract : This memoir summarises my contribution to the problems of representing and comparing the shapes and chemical properties of molecules using novel fast Fourier transform (FFT) techniques. The three main application areas considered are clustering and classifying the shapes of large protein molecules, calculating how pairs of protein molecules fit together, or "dock", to form a macromolecular complex, and comparing rapidly the shapes of many small molecules in so-called virtual drug screening. From a computational point of view, the main theme of my work is that comparing the complex three-dimensional (3D) shapes of molecules is largely a rotational problem, and therefore that molecules should be represented using polar coordinate systems in order to be able to compare them efficiently using rotational FFT correlations. Although FFT-based techniques are widely used in many areas of science, conventional Cartesian grid-based FFT approaches can accelerate shape-matching calculations in only three of the six rigid body degrees of freedom. Here, I show that by representing molecules using orthogonal expansions of spherical harmonic and Gauss-Laguerre polynomials, and by using only standard techniques of calculus, their shapes may be both rotated and translated analytically, and pairs of shapes may be compared or docked very efficiently using a series of 1D, 3D, or even 5D rotational FFTs. Although much of the underlying mathemetical theory is "well known" in the domains of theoretical chemistry and nuclear physics, the overall approach is novel in the context of molecular shape matching and protein docking, and in 3D object recognition in general. The final part of the memoir considers some future perspectives for extending the approaches developed and using them to tackle current challenges in structural systems biology such as high-throughput virtual drug screening, modelling protein flexibility during docking, and assembling very large macromolecular structures.
Document type :
Habilitation à diriger des recherches
Complete list of metadatas
Contributor : David Ritchie <>
Submitted on : Friday, April 22, 2011 - 9:02:36 AM
Last modification on : Tuesday, September 24, 2019 - 1:12:26 AM
Long-term archiving on: : Saturday, July 23, 2011 - 2:27:20 AM


  • HAL Id : tel-00587962, version 1



David Ritchie. Algorithmes Haute-Performance pour la Reconnaissance de Formes Moléculaires. Informatique [cs]. Université Henri Poincaré - Nancy I, 2011. ⟨tel-00587962⟩



Record views


Files downloads