A statistical methodology to select covariates in high-dimensional data under dependence. Application to the classification of genetic profiles associated with outcome of a non-small-cell lung cancer treatment
Résumé
We propose a new methodology to select and rank covariates associated to a
variable of interest in a context of high-dimensional data under dependence
but few observations. The methodology imbricates successively clustering of
covariates, decorrelation of covariates using Factor Latent Analysis, selection
using aggregation of adapted methods and finally ranking. Simulations study
shows the interest of the decorrelation inside the different clusters of covariates.
The objective of our method is to determine profiles of patients linked with
the outcome of a treatment. We apply our method on transcriptomic data of
n = 37 patients with advanced non-small-cell lung cancer, who have received
chemotherapy. The survival time of these patients being known, we apply our
method to select the covariates that are the most linked with the outcome
treatment among a set of more than 50 000 transcriptomic covariates. We
obtain different transcriptomic profiles for the patients whose survival time was
short, versus the other patients with longer survival time.
Origine : Fichiers produits par l'(les) auteur(s)