Consensus Molecular Subtyping of Pancreatic Ductal Adenocarcinoma and Tumor-intrinsic Regulatory Network Inference from Deconvolved Gene Expression Profiles

Description

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal diseases amongmajor malignancies, with only six-month overall survival from diagnosis. Due to limitedresponse to adjuvant chemotherapies, surgical excision provides the best chance forlonger survival. Local and distal recurrence, however, are found in 80% patients aftersurgery, precluding curative resection in more advanced disease. At present clinical,radiological, and pathological data are used for decision-making, but thesecharacteristics are insufficient to identify clinically relevant subgroups. Furthermore,genetic analysis of PDAC intertumor heterogeneity has identified many geneticalterations, but relatively few consistent driver genes. How to stratify PDAC patientsinto molecularly distinct subgroups, in relation to clinical outcomes, is critical forselection of patients for optimized adjuvant therapies and design of targeted agents.Gene expression-based subtyping has been widely accepted as a relevant source ofdisease stratification. Four major classification systems for PDAC have been reported inthe literature, demonstrating that similar to many other types of cancers PDAC is not asingle disease. Our recent work on in-house-generated RNA-Seq data for 90 stage IIPDAC patients also identified 4 molecularly distinct subtypes (PDACS1-4). Morespecifically, PDACS1 and PDACS4 are both enriched for diabetes mellitus (DM) asshown by the increased expression ofb-cell and insulin secretion signatures. However,PDACS4 shows higher immune activities, which explains its better overall survival thanPDACS1. PDACS2 shows dysregulation of various metabolism programs, mitochondrion-relatedpathways, DNA repair and MYC- and Wnt-pathways. PDACS3, with the poorestoverall survival, is characterized by upregulation of epithelial-to-mesenchymaltransition, cell migration, TGF-b1 signaling, stroma infiltration, as well as highlyexpressed PD-L1 and PD-L2 genes.Despite these various molecular subtyping effort, its translational potential is hamperedby discrepant results, which are probably due to differences in bioinformatic algorithms,diverse discovery cohorts, sample preparation methods and gene expression platforms. Ofutmost importance is to elucidate the interrelations between the five PDAC taxonomies,including ours, and define common disease patterns (or consensus molecular subtypes,CMSs) in a principled, unbiased manner, which is our 1st objective in the proposal.Through integrative analysis of multi-omic data and clinical information for ~1500PDAC samples, our 2nd objective is to comprehensively characterize identified CMSs.Focusing on the poor prognosis subtype, our 3rd objective is to dissect tumor-intrinsicregulatory mechanism based on network inference from deconvolved gene expressionprofiles, and subsequently to prioritize master regulators for experimental validation byour collaborators at the University of Amsterdam.?