Research directions

I develop methods of statistical learning to gain insights in microbial and functional ecology. In more details, here are my main research directions:

Microbial world

The microbes are the most abundant entities on Earth, encoding huge swathes of the global biodiversity. They are deeply anchored in the global nutrient cycling.

I contribute to the effort of the scientific community better understand the functions and relations at stake in these ecosystems.

Understanding functions

I have been contributing to release of various databases, either of general purposes or focussing on more specific aspects (viruses in particular). I also developed machine learning tools to unveil functions encoded in microbial genomes.

Related publications:

• Databases:
  • EnVhogDB bioRxiv
  • PHROG NAR
  • Uniclust NAR
• Tools:
  • Empathi bioRxiv
  • SPAED bioRxiv

Unveiling relations shaping microbial ecosystems

I give a particular focus on discovering the relations between bacteria and their viruses (aka bacteriophages or simply phages). Indeed, phages in relation to their host shape the structure of microbial ecosystems. On a more applied perspective, knowing which phage target which bacteria have strong clinical implications, as it could help to better handle bacterial infections through phage therapy, a promising alternative to traditional antibiotics.

Related publications:

• WIsH Bioinformatics
• SpacePHARER Bioinformatics

Macro ecology

I am interested in gaining insights about why a species establishes itself in a given environment. More specifically, I worked in collaboration with the LECA on plants, quantifying the direct effects of environment on the plant’s traits and the indirect effect mediated by the species. We also improved the prediction of plant communities by constraining on community functionnal indices (e.g. measure of mean and variance of a trait). We are in the process of integrating migration and interaction of species to explain where species establish.

Related publication:

• PFAL-AFPL MEE

Environment

Ecological models call for precise environmental inputs, with a fine spatial resolution and/or an accurate future prediction under a given climatic scenario. I worked with colleagues of IGE and MeteoFrance to train statistical models to infer accurate environmental data.

Related publications:

• The Cryosphere
• Nat. Comm.
• AIES
• NPG

Statistics and ML

Concrete questions rise more fundamental ones calling for theoretical developments.

For instance, to assess the statistical significance of spatial associations in botanical surveys, we had to develop a new statistical test on stratified contingency tables. The scope of application of this test turns out to be quite borad: epidemiology, meta-analysis, etc.

Related publication

• GASTE bioRxiv

I also provided a theoretical justification to the convergence of iterative reinjection of random noise in autoencoders:

• ICMLA