In December 2004, a call for proposals open to all scientific and medical teams was launched according to the usual AFM proceedings: postings, mass mailing, and relay of the call by the CNRS documentation and information departments.
In January 2005, 17 letters of intent were received, and ten detailed research projects were submitted in April. An expertise, independent of each project, by two biologists and two bioinformatic specialists or grid computing specialists allowed to select three new projects. A funding agreement for these projects was granted after approval by the Decrypthon board of directors, the AFM Scientific Board and the AFM Administrative Board.

The three selected projects:
Alessandra Carbone from the Université Pierre et Marie Curie, Paris (Inserm U511 –Cellular and Molecular Immunology of Parasitic Infections – Analytical Genomics)
Project coordinated by Christophe Pouzat from the Université René Descartes, Paris V (CNRS UMR 8118 – Cerebral Physiology Laboratory)
Project coordinated by Marc Robinson-Rechavi from the School of Biology and Medicine of the Université de Lausanne » (Ecology and evolution department)

These three projects started using the university Decrypthon grid at the end of 2005, by sharing the computational power thus generated.

“Large scale investigation of protein-protein, DNA-protein and protein-ligand interactions in the search of new therapeutic targets”.
 
This project aims to develop computing tools able to locate on the surface of proteins, interaction sites with DNA, ligands, and other proteins.
This development brings together the "Joint Evolutionary Trees" method (inspired by the "Evolutionary Trace" method) and the "Molecular docking" method.
The first method is based on the analysis of the evolution of a protein sequence to detect the parts of a protein that are likely to be of biological importance and therefore conserved during evolution. The second method consists of studying the interactions of two 3D protein models, in order to find the spatial positions with correct interactions. In terms of computation time, the algorithms used are very expensive and if hundreds or thousands of proteins need to be analysed the time can amount to centuries. When only the most relevant parts of the protein have been screened, the computation time becomes feasible with a computing power such as Decrypthon's.

2ptc protein complex: bovine trypsin protease and inhibitor.

Potential map, obtained after docking of the inhibitor on trypsin. (theta and phi angles locate the position of the ligand around trypsin), the ligand position in the crystallographic complex is in the middle of the picture (and corresponds to an energetic minimum).

Interaction sites between the protease and its inhibitor, detected by Joint Evolutionary Trees.

On the left, extracellular recording with a multi-electrode in the olfactory system of the insect (here the American locust, Schistocerca americana). On the right, an example of the recording obtained on four neighbouring electrodes. What interests us on these recordings are the peaks, due to the activity of the neurons from the tissue. These peaks are the images of the action potentials emitted by the neurons. In this recording, many neurons are active and the first step of the analysis of these experiments is sorting the action potentials, i.e. separation of mixed signals from raw data into a single sequence of action potentials corresponding to each active neuron in the recording.