Unsupervised subtyping and methylation landscape of pancreatic ductal adenocarcinoma

Abstract

Pancreatic Ductal Adenocarcinoma (PDAC) is an aggressive form of pancreatic cancer that typically manifests itself at an advanced stage and does not respond to most treatment modalities. The survival rate of a PDAC patient is less than 5%, with a median survival of just a couple of months. A better understanding of the molecular pathology of PDAC is needed to guide research for the development of better clinical treatment modalities for PDAC patients. Gene expression studies performed to date have identified different subtypes of PDAC with prognostic and clinical relevance. Subtypes identified to date are highly heterogeneous since pancreatic cancer is heterogeneous cancer. Tumor microenvironment and stroma constitute a major chunk of PDAC and contribute to the heterogeneity. Better subtyping methods are need of the hour for better prognosis and classification of PDAC for future personalized treatment. In this work, we have performed an integrated analysis of DNA methylation and gene expression datasets to provide better mechanistic and molecular insights into Pancreatic cancers, especially PDAC. The use of varied and diverse datasets has provided valuable insights into different cancer types and can play an integral role in revealing the complex nature of underlying biological mechanisms. We performed subtyping of TCGA-PAAD gene expression and methylation datasets into different subtypes using state-of-the-art normalization methods and unsupervised clustering methods that reveal latent hidden factors, leading to additional insights for subtyping. Differential expression and differential methylation were performed for each of the subtypes obtained from clustering. Our analysis gave a consensus of five cluster solution with relevant pathways like MAPK, MET. The five subtypes corresponded to the tumor and stromal subtypes. This analysis helps in distinguishing and identifying different subtypes based on enriched putative genes. These results help propose novel experimentally-verifiable PDAC subtyping and demonstrate that using varied data sets and integrated methods can contribute to disease prognostication and precision medicine in PDAC treatment.

Pan-Cancer DNA Methylation Analysis and Tumor Origin Identification of Carcinoma of Unknown Primary Site Based on Multi-Omics

“…In many cases, the overlap of methylation characteristics between any two cancer subtypes was significant, even though the overlap itself involved a small number of CpG sites (e.g., on the order of 25%). In addition, we also found that a set of 730 CpG sites located within the epi-driver genes had the same direction of change for three or more cancer subtypes, including genes that had previously proved to have biological functions in cancer progression such as EGFR ( Forloni et al, 2016 ; Apicella et al, 2018 ), NCOR2 ( Bhasin et al, 2015 ), MAML3 ( Sanchez-Vega et al, 2018 ), TSC2 ( Papp et al, 2018 ), and FGFR2 ( Roy et al, 2021 ) ( Supplementary Table S3 ). We further examined whether any of the CpG sites located in epi-driver genes would be significantly enriched in tumor tissue-specific gene markers associated with the given cancer subtype [as obtained using The Network of Cancer Genes (NCG)] ( Repana et al, 2019 ).…”

Epigenetic Alterations in Pancreatic Cancer Metastasis

“…These clusters were each enriched with a different tumor grade, indicating that DNA methylation can be used to estimate the histopathological stage of PDA tumors [61]. In an analysis of both TCGA-PAAD transcriptome and DNA methylome data, unsupervised subtyping of TCGA-PAAD samples based on genes whose expression was significantly correlated with methylation expression patterns was performed [62]. Interestingly, this analysis identified five subtypes, four of which correspond to the molecular subtypes identified by Bailey et al (i.e., squamous, pancreatic progenitor, immunogenic, and aberrantly differentiated endocrine exocrine [ADEX]), and the last unique subtype was enriched for tumor microenvironment related genes [42,62].…”

Unraveling Tumor Heterogeneity by Using DNA Barcoding Technologies to Develop Personalized Treatment Strategies in Advanced-Stage PDAC

“…The transcriptomic landscape of PDAC is not completely controlled by genomic changes, but rather by epigenetic modifications-a system that is highly dysregulated in PDAC [4,[29][30][31]. The transcriptomic PDAC subtypes described above can be explained in part by the epigenetic landscape, in particular, DNA methylation [29].…”