ENCODE recommends that each replicate of WGBS should have 30x coverage;
For DMR identification, sequencing at levels higher than 5-15x leads to wasted resources that would be better spent on an increased number of biological replicates.
If the goal is primarily to identify long DMRs with large methylation differences, 1-2x per sample is acceptable.
Quality control of enrichment
Lorenz curve can be used to evaluate the enrichment strength.
For RNA-based enrichment, I have written a simple script to normalize different expression levels using the paired Input sample. Please refer to MENG for more information.
Different genome and annotation versions
There are different reference assemblies for the same species. Be sure to use the currect one. Coordinates from different genome assemblies can be transformed using LiftOver or CrossMap.
Simple fold change will lead to enrichment of lowly expressed genes because of their larger variances. Be sure to use statistics such as negative binomial distribution in DESeq2 to find the differences.