Section: New Results

Metabolism: from enzyme sequences to systems ecology

Participants : Méziane Aite, Arnaud Belcour, Mael Conan, François Coste, Clémence Frioux, Jeanne Got, Anne Siegel, Hugo Talibart.

Modelling proteins with long distance dependencies [F. Coste, H. Talibart] [15], [30], [30]

• We proposed to use information on protein contacts to train probabilistic context-free grammars representing families of protein sequences. We developed the theory behind the introduction of contact constraints in maximum-likelihood and contrastive estimation schemes and implemented it in a machine learning framework for protein grammars. Evaluation showed high fidelity of grammatical descriptors to protein structures, improved precision in recognizing sequences and the ability to model a meta-family of proteins that could not be modeled by classical approaches [15].

• We then investigated the problem of modeling proteins with crossing dependencies. Motivated by their success on contact prediction, we propose to use Potts models for the purposes of modeling proteins and searching. We developed ComPotts a tool for optimal alignment and comparison of Potts models, enabling to take into account the coevolution of residues for the search of protein homologs [30], [40].

Large-scale eukaryotic metabolic network reconstruction [A. Siegel, C. Frioux, M. Aite, A. Belcour, J. Got, N. Théret, M. Conan] [17], [14], [38]. Metabolic network reconstruction has attained high standards but is still challenging for complex organisms such as eukaryotes.

• Large-scale eukaryotic metabolic network reconstruction: We participated to the reconstruction of a genome–scale metabolic network for the brown Algae Saccharina japonica and Cladosiphon okamuranus in order to shed light of the specificities on the carotenoid biosynthesis Pathway.

• Metabolic pathway inference from non genomic data: We designed methods for the identification of metabolic pathways for which enzyme information is not precise enough. As an application study, we focused on Heterocyclic Aromatic Amines (HAAs), which are environmental and food contaminants classified as probable carcinogens. Our approach based on a refinement of molecular predictions with enzyme activity scores allows to accurately predict HAAs biotransformation and their potentials DNA reactive compounds  [54].

Systems ecology: design of microbial consortia [C. Frioux, A. Belcour, J. Got, M. Aite, A. Siegel] [21], [22], [34], [33].

• We participated to the application of our methods to algal-microbial consortia, with good preliminary results, and presented them as an invited conference [22].