1. XRay crystallography 
All these structures have benefitted from excellent Xray data collection facilities in ESRF, Grenoble.
1.1 XRay crystallography and Drug Design
TMP kinase from M. tuberculosis.
The structure of thymidylate kinase TMPK which is essential for nucleotide metabolism has been solved at 1.95 Å resolution in the presence of TMP. We use this structural information to design potent new inhibitors of this target. The enzyme is active in the crystal state (see D. Bourgeois web site). 
The Trypanosoma brucei 6phosphogluconolactonase
has been crystallized and solved at 2.1 Angstrom resolution. It is a member of the PentosePhosphatePathway
(PPP) and is a good
candidate for drug design against T. brucei. This work is a collaboration with Veronique Stoven (Unite de BioInformatique Structurale) and was made possible thanks to both PT5 (P. Beguin) and PT6 (A. Haouz). 
Murine Terminal desoxynucleotidyl transferase.
Tdt is a member of the pol X family; it is responsible for the random addition of nucleotides at the N regions of immunoglobulins and Tcell receptors genes at the V(D)J junctions, during somatic recombination, thereby contributing to the generation of the diversity of the immune response (Tdt). Crystals of the catalytic domain of TdT have been obtained in the lab and the structure has been solved by MIR methods at 2.35 Å resolution. Binary complexes with either the incoming dNTP or the oligonucleotide primer have also been solved and analysed. A number of important biological implications of the structure have been found.

We reinvestigate the structural dynamics of the respective orientation of domains in the polymerase (palm, fingers and thumb) and PHP functions in bacterial polX, using both SAXS and crystallography. 
We solved the structure at 2.9 Angstrom resolution of a member of the CysLoop receptors family, in an open conformation. See Ref1, Ref2, Ref3. We are specially interested in the permeation process as well as the modulation of agonists activity by alcohol and/or anesthetics. 
2. MeanField Optimization techniques in molecular modelling and crystallography 
Mean Field optimization can be used to "decorate" very rapidly a given protein backbone with any desired protein sequence. Compared to other similar techniques, MF optimization is very fast ; it was originally developed together with P. Koehl. The output of the program are the coordinates of all the atoms of the model.
Four main applications are available :
Mean Field theory has been used to recast
in a single formalism the problem of phase optimization and phase combination, generalizing the approach
of Blow and Crick (1959) and Sim (1960) to treat rigourously density modification techniques in the
presence of an experimentally derived phase probability distribution function
in the same unifying theoretical framework.
In effect, a new statistical thermodynamics theory of
crystallographic refinement has been set up.
Because structure factors are complex numbers, field theoretic methods
had to be used to calculate the partition function Z of the system, from which the free energy can
then easily be derived. It bears strong resemble to maximum likelihood formalism but there are subtle differences.
The "best" maps are those that minimize the free energy.
Their use should increase the efficiency of automatic model building programs (Arp/wArp)
in structural genomics projects.
Click here to find out more about the derived formula giving the generalized figureofmerit and meanphase
FOM and meanphase for different cases of density modification
(use of a partial model, solvent flattening and Sayre formula) which are imposed as energy functions in reciprocal space:
definitions of Yk for these three different cases.
The formalism bears strong resemblance with maximum likelihood, but can be readily interpreted in terms of
usual thermodynamic functions such as free energy, entropy etc...
We use Mean Field formalism to write down and solve variations of the PoissonBoltzmann equation
applied to proteins in different situations (finite size of the free ions...).
We are especially interested in the case where the solvent is represented as an assembly of dipoles of variable density.
Eventually, this should have applications in drug design methodology, because it gives the solvation energy that is needed to estimate the free energy of binding of a ligand onto a macromolecule (protein).
Go to PDB_Hydro web site for a recent implementation of our PoissonBoltzmannLangevin model which treats the solvent as an assembly of selfoptimised orientable dipoles of variable density.
N.B. This site also contains a number of programs to handle, repair and refine protein coordinates (PDB) files. This is our web site for our own computational structural biology online tools.
3. Normal Modes and the Elastic Network Model: a simplified approach to structural transitions in macromolecules 
We discovered that Normal Modes are extremely good to describe structural transitions in polymerases. 