Section: Research Program

Metabolism: from enzyme sequences to systems ecology

Our researches in bioinformatics in relation with metabolic processes are driven by the understanding of non-model (eukaryote) species. Their metabolism have acquired specific features that we wish to identify with computational methods. To that goal, we combine sequence analysis with metabolic network analysis, with the final goal to understand better the metabolism of communities of organisms.

Research topics

Genomic level: characterizing enzymatic functions of protein sequences Precise characterization of functional proteins, such as enzymes or transporters, is a key to better understand and predict the actors involved in a metabolic process. In order to improve the precision of functional annotations, we develop machine learning approaches taking a sample of functional sequences as input to infer a grammar representing their key syntactical characteristics, including dependencies between residues. Our first goal is to enable an automatic semi-supervised refinement of enzymes classification [6] by combining the Protomata-Learner [50] framework - which captures local dependencies - with formal concept analysis. More challenging, we are exploring the learn of grammars representing long-distance dependencies such as those exhibited by contacts of amino-acids that are far in the sequence but close in the 3D protein folding.

System level: enriching and comparing metabolic networks for non-model organisms Non-model organisms are associated with often incomplete and poorly annotated sequences, leading to draft networks of their metabolism which largely suffer from incompleteness. In former studies, the team has developed several methods to improve the quality of eukaryotes metabolic networks, by solving several variants of the so-called Metabolic Network gap-filling problem with logical programming approaches [10], [9]. The main drawback of these approaches is that they cannot scale to the reconstruction and comparison of families of metabolic networks. Our main objective is therefore to develop new tools for the comparison of species strains at the metabolic level.

Consortium level: exploring the diversity of community consortia A new emerging field is system ecology, which aims at building predictive models of species interactions within an ecosystem for deciphering cooperative and competitive relationships between species [56]. This field raises two new issues (1) uncertainty on the species present in the ecosystem and (2) uncertainty about the global objective governing an ecosystem. To address these challenges, our first research focus is the inference of metabolic exchanges and relationships for transporter identification, based on our expertise in metabolic network gap-filling. A second very challenging focus is the prediction of transporters families by obtaining refined characterization of transporters, which are quite unexplored apart from specific databases [65].

Associated software tools

Protomata[url] is a machine learning suite for the inference of automata characterizing (functional) families of proteins at the sequence level. It provides programs to build a new kind of sequences alignments (said partial and local), learn automata and search for new family members in sequence databases. By enabling to model dependencies between positions, automata are more expressive than classical tools (PSSMs, Profile HMMs, or Prosite Patterns) and are well suited to predict new family members with a high specificity. This suite is for instance embedded in the cyanolase database  [50] to automate its updade and was used for refining the classification of HAD enzymes [6].

AuReMe workspace is designed for tractable reconstruction of metabolic networks [url]. The toolbox allows for the Automatic Reconstruction of Metabolic networks based on the combination of multiple heterogeneous data and knowledge sources [1]. The main added-values are the inclusion of graph-based tools relevant for the study of non-classical organisms (Meneco and Menetools packages), the possibility to trace the reconstruction and curation procedures (Padmet and Padmet-utils packages), and the exploration of reconstructed metabolic networks with wikis (wiki-export package, see: [url]. It also generated outputs to explore resulting networks with Askomics. It has been used for reconstructing metabolic networks of micro and macro-algae [62], extremophile bacteria [52] and communities of organisms [4].

Mpwt is a Python package for running Pathway Tools [url] on multiple genomes using multiprocessing. Pathway Tools is a comprehensive systems biology software system that is associated with the BioCyc database collection [url]. Pathway Tools is very used for reconstructing metabolic networks.

Metage2metabo is a Python tool to perform graph-based metabolic analysis starting from annotated genomes (reference genomes or metagenome-assembled genomes). It uses Mpwt to reconstruct metabolic networks for a large number of genomes. The obtained metabolic networks are then analyzed individually and collectively in order to get the added value of metabolic cooperation in microbiota over individual metabolism and to identify and screen interesting organisms among all.