Illumina assigns quite specific functional classes to the probes. 11 classes are defined (Bibikova et al. 2009 & 2011):
Relation to gene: Body, 5'UTR, 3'UTR, 1stExon, TSS1500, TSS200
Relation to CpG island: Island, N_Shelf, N_Shore, S_Shelf, S_Shore
As DMRforPairs looks at probes in close proximity to each other within each class, very specific annotations might result in too few probes per region per class and a subsequent drop in the number of identified regions and/or statistical power. This function therefore allows grouping and / or selection of classes of interest.
see description and classes_gene parameter in DMRforPairs
island_class
see description and classes_island parameter in DMRforPairs
recode
recoding scheme to use for the functional classes. Can be custom scheme (data frame) or build in scheme (0, 1 or 2). See details.
sep
Separator used in the second column of the recode parameter. Use ";" or do not specify if using the build in schemes.
Details
The recode parameter can be set to use one of the build in recoding schemes:
0: analyze all 11 classes annotated by Illumina separately
1: group Body, 5'UTR, 3'UTR into one category "gene" and TSS1500 and TSS200 in another ("tss"). All island associated classes are merged in one class "island"
2: all probes without subdivision into classes (also included probes associated with no class).
The recode parameter can also be set to a custom recoding scheme (data.frame). For example:
In this scheme the classes are merged into two categories: TSS or other gene region. Probes solely associated with CpG island-related classes are discarded. Probes not annotated to any of the 11 classes are always discarded by DMRforPairs, except when option 2 is used which collects all probes into one class (i.e. ignores classes).
If classes are unknown an m x 1 character vector with "unknown.gene" and "unknown.island" for all m rows (probes) can be used for the refgene_class and island_class parameters respectively (set "recode" to 2!).
Value
List of objects containing:
$pclass
original classes per probe (gene and island classes are merged) (m x 1 data frame)
$pclass_merged
classes after recoding (m x 1 data frame)
$no.pclass
row indexes of probes with no annotation to any of the classes specified in the recoding scheme.
$u_pclass
unique list of the classes of interest after recoding (i.e. the first column of the recode data fram)
Bibikova, M., et al., High density DNA methylation array with single CpG site resolution. Genomics, 2011. 98(4): p. 288-95.
Bibikova, M., et al., Genome-wide DNA methylation profiling using Infinium(R) assay. Epigenomics, 2009. 1(1): p. 177-200.
Examples
#merge_classes() is an integrated part of the DMRforPairs() wrapper and is
#not usually called by the user directly. Please see DMRforPairs() for
#an example.