Inference of phenotype-relevant transcriptional regulatory networks elucidates cancer type-specific regulatory mechanisms in a pan-cancer study

Amin Emad,Saurabh Sinha

doi:10.1038/s41540-021-00169-7

Abstract

Reconstruction of transcriptional regulatory networks (TRNs) is a powerful approach to unravel the gene expression programs involved in healthy and disease states of a cell. However, these networks are usually reconstructed independent of the phenotypic (or clinical) properties of the samples. Therefore, they may confound regulatory mechanisms that are specifically related to a phenotypic property with more general mechanisms underlying the full complement of the analyzed samples. In this study, we develop a method called InPheRNo to identify “phenotype-relevant” TRNs. This method is based on a probabilistic graphical model that models the simultaneous effects of multiple transcription factors (TFs) on their target genes and the statistical relationship between the target genes’ expression and the phenotype. Extensive comparison of InPheRNo with related approaches using primary tumor samples of 18 cancer types from The Cancer Genome Atlas reveals that InPheRNo can accurately reconstruct cancer type-relevant TRNs and identify cancer driver TFs. In addition, survival analysis reveals that the activity level of TFs with many target genes could distinguish patients with poor prognosis from those with better prognosis.

Highlights

Gene expression programs are responsible for many biological processes in a cell and extensive efforts have been devoted to elucidating these programs in healthy and disease states
As there are no rigorously validated metazoan Transcriptional regulatory networks (TRNs) to benchmark against, we evaluated the predicted TRNs indirectly through key transcription factors (TFs) and gene expression signatures derived from them, and showed clear improvement over several related tary Fig. 9 in Supplementary Information) to systematically combine the information on the significance of gene–phenotype associations with the information on the significance of gene–TF associations to obtain a phenotype-relevant TRN
Our results showed that the TFs with many cancer variation in the phenotypic scores/labels of samples, as depicted suggest novel drug targets or provide new insights, regarding the development and progress of cancer

Summary

INTRODUCTION

Gene expression programs are responsible for many biological processes in a cell and extensive efforts have been devoted to elucidating these programs in healthy and disease states. We believe there is a clear need for methods of TRN reconstruction that are geared towards detecting phenotype-relevant TF–gene relationships such as those idealized in Fig. 1b and 1c Such methods will draw our attention to regulatory networks that control the variation of phenotypic scores/labels among different samples (e.g., case vs control, subtypes of cancer, IC50 drug response values). We report here a new computational method called InPheRNo (Inference of Phenotype-relevant Regulatory Networks) to reconstruct TRNs that help explain the variation in the phenotypic labels/scores of samples It models the simultaneous effect of npj Systems Biology and Applications (2021) 9. These include (1) utilization of a multivariable Elastic Net model to relate the expression of multiple TFs to the expression of the target gene in the TF selection step, (2) obtaining a pseudo p value for each TF–gene pair using a multivariable OLS model, which includes the expression of all selected TFs, and (3) the design of the PGM such that for each gene it models the relationship of observed data to the latent variables representing all selected TFs simultaneously

RESULTS

METHODS

CODE AVAILABILITY