Investigating gene expression array with outliers and missing data in bladder cancer
Résumé
In this article, we present a methodology to perform selection among genes based on their expression in various groups of patients, in order to find new genetic markers for specific pathologies. Our approach is based on clustering the denoised data and computing a LASSO (Least Absolute Shrinkage and Selection Operator) estimator, in order to select the relevant genes. This latter belongs to the class of penalized regression estimators where the penalty is a multiple of the ℓ1-norm of the regression vector. Gene markers of the most severe tumor state are finally provided using the proposed approach.