Testing for pathway (in)activation by using Gaussian graphical models

Wessel N. van Wieringen*, Carel F.W. Peeters, Renee X. de Menezes, Mark A. van de Wiel

*Corresponding author for this work

Research output: Contribution to JournalArticleAcademicpeer-review


Genes work together in sets known as pathways to contribute to cellular processes, such as apoptosis and cell proliferation. Pathway activation, or inactivation, may be reflected in varying partial correlations between the levels of expression of the genes that constitute the pathway. Here we present a method to identify pathway activation status from two-sample studies. By modelling the levels of expression in each group by using a Gaussian graphical model, their partial correlations are proportional, differing by a common multiplier that reflects the activation status. We estimate model parameters by means of penalized maximum likelihood and evaluate the estimation procedure performance in a simulation study. A permutation scheme to test for pathway activation status is proposed. A reanalysis of publicly available data on the hedgehog pathway in normal and cancer prostate tissue shows its activation in the disease group: an indication that this pathway is involved in oncogenesis. Extensive diagnostics employed in the reanalysis complete the methodology proposed.

Original languageEnglish
Pages (from-to)1-18
Number of pages18
JournalJournal of the Royal Statistical Society. Series C: Applied Statistics
Issue number5
Publication statusPublished - 16 Apr 2018


  • Conditional independence
  • Multivariate normality
  • Network
  • Partial correlations
  • Penalized estimation


Dive into the research topics of 'Testing for pathway (in)activation by using Gaussian graphical models'. Together they form a unique fingerprint.

Cite this