Fig. 1

Identification of network biomarkers in EAC. a Using the TCGA DNA methylation and mRNA expression datasets for EAC, we identify gene-modules of joint epigenetic and expression deregulation in EAC compared to normal-adjacent tissue, using our Functional Epigenetic Modules (FEM) algorithm whilst adjusting for stromal heterogeneity. The latter is accomplished by estimating total epithelial, total immune cell and total fibroblast fractions in the TCGA samples using our HEpiDISH algorithm. FEM searches for gene-modules in the context of a PPI network. Subsequently, we apply our CellDMC algorithm to ascertain if the DNA methylation changes underlying the inferred gene modules are happening in the epithelial compartment of the tissue. Finally, we validate inferred modules in independent DNA methylation and mRNA expression EAC datasets. b Modules validating in (a) are then explored for their potential utility as early detection markers. This is done in two ways. In one case we analyse scRNA-Seq data from Barrett’s and normal esophagus to explore if any modules are deregulated in Barrett’s. In the second case, we explore if promising gene-modules exhibit variable DNA methylation patterns in saliva from cohorts containing both EAC and healthy subjects, where we also apply cell-type deconvolution methods to estimate epithelial fractions in saliva