I use logic programming to study the metabolism of non-model organisms: facilitate the refinement of metabolic networks, propose solution to select communities of interest within microbiotas.
I focus my research on graph-based or hybrid (graph-based and constraint-based) modeling of metabolism based on Answer Set Programming ↗ and the solving of combinatorial optimization problems. I contributed to the meneco ↗ project, a tool dedicated to gap-fill genome-scale metablic networks by selecting a minimal number of reactions from a database, in order to ensure the producibility of compounds of interest (PLOS Computational Biology 2017). I then collaborated to the creation of a hybrid method for metabolism gap-filling, enabling to optimize not only the constraint-based producibility of metabolites, but also the reaction of interest with the Flux Balance Analysis framework (LPNMR Conference 2017, best Student Paper). I now focus on the selection of communities within large microbiotas.
Investigating host-microbiota cooperation with gap-filling optimisation problems
PhD defense on November 19th, 2018 at Inria-IRISA Rennes.
Systems biology relies on computational biology to integrate knowledge and data, for a better understanding of organisms’ physiology. Challenges reside in the applicability of methods and tools to non-model organisms, for which data is limited, and more generally to host-microbiota systems. Understanding the interactions in the later is an objective of systems ecology. Metabolic networks are a useful solution to model them functionally. In this direction, several semantics exist and are at the core of metabolic network reconstruction, particularly for their refinement through gap-filling. Gap-filling is a combinatorial problem that aims at selecting reactions in databases to ensure the feasibility of a behaviour by the model. It is a very crucial step due to various pitfalls: model overfitting, false positive, choice of functionality semantics. This thesis aimed at better understand these limits and propose solutions to them. As a first results part, we benchmarked several gap-filling algorithms to assess the value of graph-based semantics with respect to the constraint based one. Then we propose a hybrid gap-filling method that reconciles both semantics. Finally, we extended the gap-filling problem towards the selection of communities and the screening of metabolic functions within large microbiotas. Problems modelled and solved during this PhD were applied to brown algae metabolism and to the human gut microbiota.
Here is some computational biology software I developed or for which I was involved in the development:
Some conference references and/or pdf may be missing.