hypermr - Detecting hypermethylated regions

Synopsis

$ dnmtools hypermr [OPTIONS] <input.meth>

Description

The plant genomes, exemplified by A. thaliana, are devoid of DNA methylation by default, with genic regions and transposons being hyper-methylated, which we termed HyperMRs to stress their difference from hypo-methylated regions in mammalian methylomes. DNA methylation in plants has been associated with expression regulation and transposon repression, and therefore characterizing HyperMRs is of much biological relevance. In addition to plants, hydroxymethylation tends to appear in a small fraction of the mammalian genome, and therefore it makes sense to identify hyper-hydroxymethylated regions.

The first kind of HyperMR analysis involves finding continuous blocks of hyper-methylated CpGs with the hmr program. Since the hmr program is designed to find hypo-methylated regions, one can use it to identify HyperMRs by inverting the methylation levels in the methcounts output file as follows:

$ awk '{$5=1-$5; print $0}' input.meth > input_inverted.meth

Next one may use the hmr program to find "valleys" in the inverted Arabidopsis methylome, which are the hyper-methylated regions in the original methylome. The command is invoked as below

$ dnmtools hmr -o output.hmr input_inverted.meth

This kind of HyperMR analysis produces continuous blocks of hyper-methylated CpGs. However in some regions, intragenic regions in particular, such continuous blocks of hyper-methylated CpGs are separated by a few unmethylated CpGs, which have distinct sequence preference when compared to those CpGs in the majority of unmethylated genome. The blocks of hyper-methylated CpGs and gap CpGs together form composite HyperMRs. The hypermr program, which implements a three-state HMM, is used to identify such HyperMRs. Suppose the counts output file is Col0 Meth.bed, to find HyperMRs from this dataset, run

$ dnmtools hypermr -o output.hypermr input.meth

The output file is a 6-column BED file. The first three columns give the chromosome, starting position and ending position of that HyperMR. The fourth column starts with the hyper:, followed by the number of CpGs within this HyperMR. The fifth column is the accumulative methylation level of all CpGs. The last column indicates the strand, which is always +.

Lastly, it is worth noting that plants exhibit significantly more methylation in the non-CpG context, and therefore inclusion of non-CpG methylation in the calling of hyper-methylated regions could possibly be informative. We suggest separating each cytosine context from the methcounts output file as illustrated in the previous section (via grep) and calling HyperMRs separately for each context.

Options

 -o, -out

output BED file (default: STDOUT)

 -s, -scores

output file for posterior scores

 -t, -tolerance

tolerance (default: 0)

 -d, -desert

maximum distance between covered CpGs in HyperMR (default: 1000)

 -i, -itr

max number of iterations (default: 100)

 -V, -viterbi

Use Viterbi decoding

 -M, -min-meth

min cumulative methylation level in HypeMR (default: 4)

 -v, -verbose

print more run info to STDERR while the program is running.

 -P, -params-in

HMM parameters input file

 -p, -params-out

HMM parameters output file