|
Abstract:
|
Metagenomic analysis of the human microbiome provides a rich set of microbial features that can be used for prediction and biomarker discovery tasks in the context of health and disease. In this talk I will introduce MetAML, a machine learning-based framework for prediction tasks from shotgun metagenomics data. The framework has been developed in conjunction with curatedMetagenomicData, a Bioconductor resource that provides almost ten thousand of processed metagenomic profiles from publicly available datasets. These data are distributed to the R desktop through ExperimentHub, a convenient cloud-based package. We will show a meta-analysis conducted on hundreds of metagenomes and aimed at linking the gut microbiome with colorectal cancer. We will highlight novel biomarkers and validate them across cohorts and populations. The set of metagenomes collected in curatedMetagenomicData was also the basis of a recent work able to build a large genome catalog from the human microbiome, which highlighted an extensive diversity across ages, countries and lifestyles.
|
