Section: Partnerships and Cooperations

Regional Initiatives

PEPS Mirabelle EXPLOD-Biomed

Participants : Adrien Coulet [contact person] , Malika Smaïl-Tabbone.

This project has initiated a collaboration with geneticists from the Hospital of Nancy, namely Philippe Jonveaux and Céline Bonnet. The aim of the EXPLOD-Biomed project is to propose novel knowledge discovery methods applied to Linked Open Data for discovering gene that could be responsible for intellectual deficiencies. Linked Open Data are available on-line, interconnected and encoded in a format which can be straightforwardly mapped to ontologies. Thus they offer novel opportunities for knowledge discovery in biomedical data. Here, geneticists play the role of experts and guide the knowledge discovery process at different steps.

Hydreos

Participant : Jean-François Mari [contact person] .

Hydreos is a state organization –actually a so-called “Pôle de compétitivité”– aimed at evaluating the delivering and the quality of water (http://www.hydreos.fr/fr ). Actually, data about water resources rely on many agronomic variables, including land use successions. The data to be analyzed are obtained by surveys or by satellite images and describe the land use at the level of the agricultural parcel. Then there is a search for detecting changes in land use and for correlating these changes to groundwater quality. Accordingly, one main challenge in our participation in Hydreos is to process and analyze space-time data for reaching a better understanding of the changes in the organization of a territory.

The systems ARPEnTAge (see §  6.2.2 ) and CarottAge (see §  6.2.1 ) are used in this context, especially by agronomists of INRA (ASTER Mirecourt http://www6.nancy.inra.fr/sad-aster . Currently, various display tools are under study and implementation for providing the agrnomy expert an easier interpretation of the clustering outputs http://www.loria.fr/~jfmari/App/Arpentage/Yar.avi .

PEPS Truffinet

Participant : Chedy Raïssi [contact person] .

The Truffinet PEPS project aims at developing new graph mining methods and tools to support knowledge discovery from the truffle's complex network of interactions happening in the soil between different bacterias and the subterranean Ascomycete fungus. This work uses Log-Linear Analysis (LogLA) which is a well established statistical technique for finding associations between discrete variables in data. The general objective of LogLA is to select a model that satisfactorily explains the observed frequencies of a given categorical dataset. General approaches to LogLA are exponential with respect to the number of variables. Recently, new approaches based on multiplicative log-linear models and using notions from graph theory have been developed. We applied successfully these methods in the case of the truffle bacterial environment to discover new associations in our data.

The Truffinet PEPS project involves several partners among which Intitut Elie Cartan de Lorraine (IECL), Intitut National de Recherche en Agronomie (INRA) and Centre de Recherche en Automatique de Nancy (CRAN) along with Inria Nancy Grand Est.