Network perturbation analysis identifies three molecular subtypes with distinct clinical outcomes and therapeutic vulnerabilities in endometrial cancer (IMAGE)
Caption
(A) Workflow of network perturbation analysis using gene expression data from EC and normal tissues. Our approach leverages the principle that physiological gene regulatory networks maintain homeostatic stability, whereas pathological states induce systematic disruptions in these interaction patterns. The analytical pipeline comprised four sequential stages. Initially, we generated gene expression rankings for each sample in both EC and normal sample cohorts, constructing sample-specific rank matrices. We then mapped these rankings onto a reference interaction network derived from the Reactome database, computing differential rankings between interacting gene pairs to generate delta rank matrices. To quantify disease-specific perturbations, we established a baseline interaction profile from healthy controls and calculated deviation scores for each gene pair in EC samples. This yielded an interaction perturbation matrix that captured network-level alterations associated with EC pathophysiology. (B-C) Perturbation patterns showing clear separation between tumor and normal samples. (D) Functional enrichment of 694 perturbed genes in therapeutic resistance pathways. (E) Clustering evaluation showing optimal separation at k=3. (F) Consensus clustering heatmap. (G-H) CDF curves and PAC analysis confirming three-subtype stability. (I) PCA showing spatial segregation of three patient groups (C1, C2, C3). (J) Kaplan-Meier curves showing C1 with poorest prognosis, C2 and C3 with favorable outcomes. (K) Validation of subtypes across four cohorts using NTP algorithm with signature genes.
Credit
Hengrui Liu, University of Cambridge; Hao Chi, University of Hawaii at Manoa; Jingyuan Ning, Chinese Academy of Medical Sciences
Usage Restrictions
None
License
Original content