Séminaire de Probabilités et Statistique
lundi 19 octobre 2020 à 13:45 - UM - Bât 09 - Salle 330
Antoine Bichat (Laboratoire de Mathématiques et Modélisation d'Évry)
Hierarchical FDR correction and applications to metagenomics differentially abundance studies
In metagenomics, differential abundances studies are commonly used to identify species whose abundances are associated to a phenotype of interest, e.g. healthy versus diseased. Most state of the art methods carry out independent tests to know whether each species is differentially abundant or not. However, such procedures do not take into account evolutionary relationships between species which may result in correlated abundances and then correlated p-values. These relationships are captured by the species phylogenetic tree. This work proposes a method to detect differentially abundant species based on the phylogeny. p-values are seen as a function of an Ornstein-Uhlenbeck process with shifts on the phylogenetic tree. Our method estimates the positions and intensities of those shifts. Shifts are then propagated back to the leaves -i.e. species- with confidence intervals in order to identify the differentially abundant ones.
INSCRIPTION pour participation sur place (limité à 10 personnes, salle 330) : https://framaforms.org/seminaire-de-probabilites-et-statistiques-du-1910-1602488907
WEBINAIRE ouvert à toutes et tous : https://umontpellier-fr.zoom.us/j/85813807839