Mots-Clés

Microbial ecology, omic data, probabilistic model, soil, dimensionality reduction, climate change

Description

Title

Master internship: “Modeling the response of soil microbial communities to climate change through dimensionality reduction of ‘omic data”

Host team

Équipe-projet Pléiade Inria-INRAE, Centre Inria de l’Université de Bordeaux

Supervision

Guilhem Sommeria-Klein, Inria researcher

Internship duration and dates

Flexible; expected duration and dates are 6 months between January and June

Context

The rise of “‘omic” data allows for a more and more thorough and comprehensive characterization of microbial communities (Taş et al., 2021). The amplification and sequencing of “barcode” sequences allows us to identify the microbial taxa present in the community after comparison with a reference database, and assess their relative abundances (metabarcoding). The bulk amplification and sequencing of DNA fragments allows us to reconstruct the gene content of the community (metagenomics), which informs us on all potential molecular functions. Furthermore, mass spectrometry allows the identification of the metabolites found in the community and their abundance (metabolomics), which informs us on the metabolic activity of the community and thus the molecular functions effectively performed at the time of sampling.

These data can potentially be collected from any naturally-occurring microbial community amenable to sampling. However, their processing and interpretation present challenges, which generally increase with the diversity of the sampled community. One approach to facilitate the interpretation of ‘omic data and their incorporation into predictive models is dimensionality reduction. In particular, probabilistic models based on Latent Dirichlet Allocation have been shown to provide a robust way to decompose microbial communities into assemblages of taxa that tend to co-occur and covary in abundance across samples (Sommeria-Klein et al., 2020, 2021). Similar approaches can be applied to model groups of genes from metagenomic data (Labarthe et al., 2023).

Soils harbor the most diverse and poorly known of microbial communities (Fierer, 2017). These communities ensure the recycling of plant organic matter and are therefore key to the productivity of terrestrial ecosystems, including forests and agricultural land. Their activity also controls the storage of carbon in the ground. To understand the response of soil communities to climate change, our collaborators are experimentally submitting soil collected from different environments to varying levels of heat and drought, monitoring their response through metabarcoding, metagenomics and metabolomics.

The internship is offered as part of the MISTIC project funded by the PEPR “Agroecology and ICT”. MISTIC aims at designing new multi-omic data analysis tools and developing multi-scale spatio-temporal models of microbial communities that are able to decipher the links between community structure and biological functions.

Objectives

The general objective of the research is to explore approaches of dimensionality reduction that can summarize the composition and metabolic activity of soil microbial communities in a way that is robust and predictable given environmental conditions, with the eventual goal of informing climate models. To do this, the intern will rely on previous work by the supervisor on probabilistic models of dimensionality reduction for soil microbial communities (Sommeria-Klein et al., 2020) and on soil microbial community data from MISTIC project partners and other collaborators (see Sorensen & Shade, 2020; Bandopadhyay et al., 2024 for examples). The initial focus will be on metabarcoding and metagenomic data and on bacteria for the internship, with possibility to move on to joint modelling of metabarcoding, metagenomic and metabolomic data according to progress (see Argelaguet et al., 2018 for an example). The internship is intended to lead to a PhD funded by the MISTIC project, if there is mutual interest.

Tasks

The first step will be to expand an existing Stan implementation of Latent Dirichlet Allocation by the supervisor to include covariates, following the example of Shimizu et al. (2023). This will allow us to decompose soil microbial community composition into assemblages of co-occurring microbial taxa whose niche is constrained by environmental covariates, or alternatively, groups of genes whose presence is linked to environmental covariates. The second step will be to use the developed model to search for a community decomposition that maximizes the predictability of the community response to heat and drought. Nevertheless, there will be flexibility in choosing the best modelling approach to the problem.

Working environment

The intern will join the Pléiade research group at the Inria Centre of the University of Bordeaux. The team hosts researchers from both Inria and INRAE institutes, with expertise spanning over computer science, mathematics, ecology and bioinformatics. The working language will be French or English.

Internship allowance of 4,35€/hr. 

Benefits include:

• Subsidized meals
• Partially reimbursed public transportation

Requirements

The intern is expected to at least be familiar with the mathematical notations used in statistics and to be proficient in at least one interpreted coding language (e.g., R or Python). Prior experience with Bayesian inference and the use of a probabilistic programming language (such as Stan) as well as background knowledge of microbial ecology is appreciated. Furthermore, good writing and communication skills will be essential for pursuing towards a PhD.

References

Argelaguet, R., Velten, B., Arnol, D., Dietrich, S., Zenz, T., Marioni, J.C., Buettner, F., Huber, W. & Stegle, O. (2018) Multi-Omics Factor Analysis—a framework for unsupervised integration of multi-omics data sets. Molecular Systems Biology, 14, e8124.

Bandopadhyay, S., Li, X., Bowsher, A.W., Last, R.L. & Shade, A. (2024) Disentangling plant- and environment-mediated drivers of active rhizosphere bacterial community dynamics during short-term drought. Nature Communications, 15, 6347.

Fierer, N. (2017) Embracing the unknown: disentangling the complexities of the soil microbiome. Nature Reviews Microbiology, 15, 579–590.

Labarthe, S., Plancade, S., Raguideau, S., Plaza Oñate, F., Le Chatelier, E., Leclerc, M. & Laroche, B. (2023) Four functional profiles for fibre and mucin metabolism in the human gut microbiome. Microbiome, 11, 231.

Shimizu, G.Y., Izbicki, R. & Valle, D. (2023) A new LDA formulation with covariates. Communications in Statistics - Simulation and Computation, 0, 1–18.

Sommeria-Klein, G., Watteaux, R., Ibarbalz, F.M., Pierella Karlusich, J.J., Iudicone, D., Bowler, C. & Morlon, H. (2021) Global drivers of eukaryotic plankton biogeography in the sunlit ocean. Science, 374, 594–599.

Sommeria-Klein, G., Zinger, L., Coissac, E., Iribar, A., Schimann, H., Taberlet, P. & Chave, J. (2020) Latent Dirichlet Allocation reveals spatial and taxonomic structure in a DNA‐based census of soil biodiversity from a tropical forest. Molecular Ecology Resources, 20, 371–386.

Sorensen, J.W. & Shade, A. (2020) Dormancy dynamics and dispersal contribute to soil microbiome resilience. Philosophical Transactions of the Royal Society B: Biological Sciences, 375, 20190255.

Taş, N., de Jong, A.E., Li, Y., Trubl, G., Xue, Y. & Dove, N.C. (2021) Metagenomic tools in microbial ecology research. Current Opinion in Biotechnology, 67, 184–191.