Abstract
Previous studies have generated controversial findings regarding the correlation between DNA methylation in the human genome and gene expression. Some reports have indicated that promoter methylation is negatively correlated with gene expression levels; however, in some cases, a poor or positive correlation was reported. Most previous findings were based on general trends observed with whole-genome data analysis. Here, we present a novel chromosome-specific statistical analysis design of empirical Bayes differential tests for five phases of erythroid development. To better understand the common methylation patterns of differentially methylated regions (DMRs) during specific stages, we defined differential phases for each CpG locus, based on a maximum log2 fold change. Analyzing hypermethylated and hypomethylated CpG loci separately showed variations in methylation patterns during erythropoiesis in the gene body, promoter and enhancer regions. Hypomethylated DMRs showed stronger associations with erythroid-specific enhancers at the differentiation start phase and with exons in the intermediate phase. To investigate the hypomethylated DMRs further, transcription factor binding site-enrichment analysis was conducted. This analysis highlighted novel transcription factors during each differentiation stage that were not detected by previous differential methylation data analysis. In contrast, hypermethylated DMRs showed a consistent methylation pattern over the different genomic regions. Thus, a closer examination of DNA methylation patterns in a single chromosome during each developmental stage can contribute to verify the association nature between gene expression and DNA methylation.