Table 1.

Epigenome-wide association studies (EWAS) identifying methylation markers associated with pancreatic ductal adenocarcinoma (PDAC).

Source	Tissue Type	Techniques	Sample Size	Comparisons	Strengths	Weakness
[123]	Pancreas tissue	88K Agilent promotor array and 244K island array—methylated CpG island amplification (MCA)	10 pancreatic cancer cell lines; normal human pancreatic ductal epithelium (HPDE) and human pancreatic Nestin-expressing cells (HPNE)	Cancer vs. normal	● Study conducted in cells lines and patient tissue	● Early implementation of technology
			57 pancreatic cancer samples and 34 normal pancreas samples		● Investigated using several approaches	● Limited number of probes
[124]	Leukocytes	Illumina GoldenGate methylation Beadchip—1505 CpG sites	Phase 1: 132 never-smoker cases and 60 never-smoker controls	Cancer vs. normal	● Validation	● Limited number of probes
			Phase 2: 240 cases and 240 matched controls (half never smokers)		● Adjustment for some confounders	● Promotor region CpGs only
[54]	Cell lines and pancreas tissue samples	244K ChIP-on-Chip microarray—27800 CpG island array	9 pairs of pancreatic cancer versus normal pancreatic epithelial tissues	Cancer vs. normal	● Looked at number different cell lines and tissue samples	● CpG islands only
			3 matched pairs of pancreatic cancer versus lymphoid tissue from same individual		● No validation within this study	● Looked at methylation difference as individual samples rather than average of population
[125]	Pancreas tissue samples	Methyl capture sequencing method—(methylCap-Seq)	10 pancreatic cancer tissues and 10 adjacent non-tumor tissues	Cancer vs. normal	● Explored potential functional result of CpG methylation	● Used p-value < 0.05
					● 728/3911 differently methylated genes identified that were also reported in at least one of 3 different studies	● Early implementation of technology
[92]	Pancreas tissue samples	Infinium 450k methylation array (Illumina)	167 untreated PDACs and 29 adjacent normal pancreata	Cancer vs. normal	● Larger sample size	● No discussion of the significance of dissimilar pathway analysis results using two different methods
			121 PDAC and 8 nontumor	Survival	● Looked at methylation across potential confounding factors	● Survival analysis methods not described
					● 850/3522 genes previously reported to have differential methylation
					● Determined significance based on p-value and beta value
[126]	Pancreas tissue samples	HumanMethylation450k Beadchip (Illumina)	Secondary analysis of public TCGA data - 184 tumors and 10 normals	Cancer vs. normal	● Looked at methylation and expression, as well as mutation loads and copy number variations of key oncogenes or suppressor genes	● Had to attempt to adjust for batch effects using PCA
					● Promoter region methylation highly negatively correlated with gene expression	● Used median beta value for genes with multiple methylation markers with no justification
					● Non-promoter region methylation highly positively correlated with gene expression	● Stated gender bias was ignored by excluding X and Y chromosomes
					● Determined significance based on p-value and beta value	● Used only beta value for significance
					● Highlighted methylation of genes coding for other epigenetic markers
[127]	PDX – pancreas tissues - stem cells	HumanMethylation450k Beadchip (Illumina)	Not given	Cancer stems cells vs. non-cancer stem cells	● Looked at stem cells from PDAC-185, liver met (PDAC-A6L) and single cell-derived tumor	● Unknown systematic effects of DNMT1 treatment
					● Function of stems cells reduced by inhibiting DNMT1	● Unknown sample size used
					● Cancer stem cells show hypermethylation in intergenic regions
[128]	PDX – pancreas tissue	Chip-seq	24 xenograft samples - tumor	Survival	● Looked at chromatin states, DNA methylation, Gene expression, and Transcription factors	● limited to later stage samples
		RNA-seq
		MethEpic