Incorporation of gene exchangeabilities improves the reproducibility of gene set rankings

Charlotte Soneson, Magnus Fontes

Research output: Contribution to journalArticlepeer-review

Abstract

Gene set-based analysis methods have recently gained increasing popularity for analysis of microarray data. Several studies have indicated that the results from such methods are more reproducible and more easily interpretable than the results from single gene-based methods. A new method for ranking gene sets with respect to their association with a given predictor or response, using a new framework for robust gene list representation, is proposed. Employing the concept of exchangeability of random variables, this method attempts to account for the functional redundancy among the genes. Compared to other evaluated methods for gene set ranking, the proposed method yields rankings that are more robust with respect to sampling variations in the underlying data, which allows more reliable biological conclusions. (C) 2012 Elsevier B.V. All rights reserved.
Original languageEnglish
Pages (from-to)588-598
JournalComputational Statistics & Data Analysis
Volume71
DOIs
Publication statusPublished - 2014

Subject classification (UKÄ)

  • Mathematics

Keywords

  • Bioinformatics
  • Exchangeability
  • Gene set ranking
  • Microarray

Fingerprint

Dive into the research topics of 'Incorporation of gene exchangeabilities improves the reproducibility of gene set rankings'. Together they form a unique fingerprint.

Cite this