Finding differentially methylated regions
It is possible to run the Call Methylation Levels tool (see http://resources.qiagenbioinformatics.com/manuals/clcgenomicsworkbench/current/index.php?manual=Call_Methylation_Levels.html) to detect regions of differential methylation after the Detect QIAseq Methylation workflow has been run.
For detecting Differentially Methylated Regions using QIAseq Targeted Methyl data, we advise using certain settings of the Call Methylation Levels tool:
- Uncheck Ignore duplicate matches: Duplicate matches have already been identified in the Detect QIAseq Methylation workflow using UMIs.
- Uncheck Confirm methylation contexts in reads: this option, when enabled, discards reads where the context is different in the reads and reference, for example due to a SNV. When this is disabled, SNVs that change the methylation context (for example from CpG to CHG) cannot be distinguished from changes in methylation. However this is unlikely to lead to incorrect interpretation of the data. The presence of a SNV can be confirmed by inspection in the read mapping.
- Specify the target regions file of the workflow in the field Restrict calling to target regions: Providing this will report whether each target is differentially methylated. This makes sense if the targets are the unit of biological interest, or if they are short enough that methylation patterns within a target are expected to be constant i.e. most Cs are hypomethylated/hypermethylated/unchanged when compared to a control sample. Note that setting this file means that the "window length" parameter is ignored.
- Set the Statistic mode to ANOVA
- Set the Minimum high-confidence site coverage to 10 or more: Coverage should be high for most targets, so it may be acceptable to exclude low coverage sites which will typically be far from the primers, and which may add more noise than signal. However, first check that this is higher than the typical coverage of positions in each target using the Final_target_coverage output.
- Uncheck Create track of methylated cytosines: this would generate the same per-sample data already produced in the workflow.