Utilisation du microbiome intestinal dans la prédiction de l'état de santé de l'hôte
|Advisor:||Raymond, Frédéric; Di Marzo, Vincenzo|
|Abstract:||During the last decades, research positioned the gut microbiome as a major regulator of numerous physiological processes in humans. Propelled by next-generation sequencing technologies, the research on microbial ecology has undergone a significant paradigm shift; generally, bacteria isolation and cultivation are now being replaced by genetic sequencing of whole bacterial communities directly from their environment. This type of analysis, referred as metagenomics, revealed the large catalog of microbial genes comprised in the gut environment and lifted the veil on the microbiota's silent majority: non-cultivable microorganisms. This vast catalog of genes represents a real mine of information in a context where research aims at finding molecular mechanisms to explain the relation between microbiome and host health. In this context, machine learning, which allows the analysis of complex data, can be used to point toward promising microbial features. The objective of this project is to use metagenomics data from healthy and diseased individuals in a classification task. More precisely, our goal is to compare the predictive power of different microbiome representations, all derived from untargeted metagenomics data. Our study has shown that in a context of host phenotype classification, representation methods that use all the available sequenced information allow better prediction performances than those that are based on reference databases, like the taxonomic and functional profiles. Our results suggest that the exclusive use of a priori information, in a machine learning context, limits, in a way, the possibility of finding new microbial effectors unknown from reference databases.|
|Document Type:||Mémoire de maîtrise|
|Open Access Date:||10 January 2022|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.