DNA methylation data generation and analysis

Whole genome DNA methylation profiling was done from the DNA isolated from the tumour and adjacent normal tissues of validation cohort 1 (n = 7) (Additional file 2: Method 2) using Infinium Human Methylation 450 K Bead Chip (Illumina). The bead chip included 485,577 CpG sites and targets 96% of CpG islands in human genome (hg19), using 4 μl of bisulfite-treated DNA, according to the manufacturer’s protocol [16]. The raw signal intensities on bead chips were read and converted to raw data (IDAT files) using the iScan platform (Illumina). The raw data including the IDAT files is available in the GEO database with the entry “{"type":"entrez-geo","attrs":{"text":"GSE181740","term_id":"181740"}}GSE181740”.

In this current study, we have followed the analysis done by Aryee et al.,2014 [16] for processing of DNA methylome data. For each CpG site in the genome, methylation level was assessed using the established β value. The β value was calculated as the ratio of fluorescent signal intensity of the methylated (M) and total of signal intensities from the methylated (M) and unmethylated (U) alleles:

We have used the minfi R package R version 4.0.4 to convert raw array data to β value and for data normalisation. To reduce false positive inferences from our data, we have used data quality control as described by Aryee et al.,2014 [16], Probes were excluded from further analysis if: (1) their detection p-value > 0.05, (2) they contain SNPs in their sequences, (3) they are positioned on X and Y chromosomes. With the filtered probe set, we performed both intra-array (Infinium I and Infinium II) normalisation and inter-array normalisation using quantile normalisation approach (using β values).

Prior to differential methylation calculation, purity of tumour samples was estimated using “Infinium Purify R” package using R version 4.0.4 [17]. Differentially methylated positions (DMPs) between the tumour and the normal tissue samples were identified by comparing the average beta values between the study groups. The probes with average |Δβ| (difference between the average β-value among the tumour samples and the average β-value among the normal samples) ≥ 0.2 and Benjamini–Hochberg corrected p < 0.05 were considered as significantly differentially methylated positions (DMPs). Hyper-methylated positions (CpG sites) were DMPs with Δβ ≥ 0.2 and hypo-methylated sites were DMPs with Δβ ≤ − 0.2. The DMPs were annotated to genes, referred to as “gDMs”.

Association of differential methylation at key genes with PanCa, was assessed using MSP in validation cohort 2 (n = 22). The band intensities after electrophoresis were noted for each MSP and the data was represented as percentage (%) methylation among all the samples for a gene. The detailed MSP procedure is mentioned in Additional file 2: Method 3.