Modeling the biodiversity patterns arising from environmental DNA surveys and application to the tropical forest of French Guiana
General patterns of biodiversity, such as the diversity, relative abundance and spatial distribution of organisms, have been the focus of much research since the onset of ecological sciences with the hope of discovering general rules governing them. Unfortunately, those patterns are still poorly known for most groups of biological organisms by lack of rapid and repeatable methods for the measurement of biodiversity. This is especially true in the tropics where most biodiversity is undescribed and where even the census of well-known organisms such as trees relies on rare and highly specialized naturalist expertise.
DNA-based methods are rapidly transforming this state of knowledge thank to the emergence of high-throughput DNA sequencing in the last ten years. In particular, extracting and sequencing the DNA contained in environmental samples such as soil or water, called ’environmental DNA’, is a promising method for conducting fast, standardized and comprehensive biodiversity surveys. As of currently, ’metabarcoding’ is the most widespread method for exploiting the information contained in environmental DNA: it consists in amplifying a short DNA barcode by PCR from the DNA contained in an environmental sample, sequencing the product by high-throughput sequencing, and clustering the resulting sequences into Molecular Operational Taxonomic Units (MOTUs), which are proxies for species.
I use both statistical and mechanistic models to reveal and understand the biodiversity patterns emerging from spatially distributed DNA-based biodiversity surveys.
Such surveys generate large datasets consisting in the abundances of tens of thousands of MOTUs in spatially distributed environmental samples. I use a data mining algorithm initially developed for classifying text documents into topics (’topic modeling’) to reveal the spatial patterns of co-occurence between MOTUs. Those patterns can then be compared to the spatial variability of environmental conditions. I apply this method to soil DNA datasets describing the spatial distribution of biodiversity across domains of life over several hectares of primary tropical forest in French Guiana.
Mechanistic models represent another approach for interpreting biodiversity patterns. Hubbell’s neutral model is the most successful attempt so far at a simple, general, mechanistic and parsimonious biodiversity model. In particular, exact maximum-likelihood parameter inference allows comparing and interpreting patterns of relative species abundances based on the values of neutral model parameters. I investigate how to fit this model to DNA-based species abundance distributions despite the differences between DNA-based data and classical biodiversity census data. I do so by taking into account the uncertainty sources associated with DNA-based biodiversity measurement.
In order to study beta-diversity patterns at different spatial scales, we have collected soil DNA data in a number of 1-ha forest plots 100 m to 100 km away from each other in French Guiana. I use statistical methods to assess the relative contribution of distance and environmental conditions to the taxonomic dissimilarity among the plots, and I compare the decay of taxonomic similarity with distance to the neutral prediction.
Aside from my PhD advisors, I much benefit from the collaboration with Lucie Zinger for the bioinformatics and data analysis parts and with Pierre Taberlet’s team (LECA, Grenoble, France) for the molecular and fieldwork parts.
Background and previous projects
In 2012-2013, I worked on modeling the propagation of neuronal activity in the cortex with Rava da Silveira and Vincent Hakim at the LPS (ENS, Paris). In 2011, I worked on modeling short-term memory in the cortex using dynamical systems with Mikhail Rabinovich and Ramon Huerta at the BioCircuit Institute (UCSD, California).
I studied as an undergraduate and took my master’s degree in physics at the Ecole Normale Supérieure de Lyon (ENS Lyon).