This function will read in Bismark coverage files and optionally filter for coverage and feature-wise variance.
Usage
read_bismark(
files,
coverage = 1,
prop_samples = 1,
remove_zero_var = FALSE,
return_mats = FALSE
)Arguments
- files
Named vector of file paths to the bismark coverage files
- coverage
Minimum coverage for a CpG site. Default (1)
- prop_samples
Proportion of samples that site must be covered to be retained. Default (1)
- remove_zero_var
Should zero variance features be removed after filtering for coverage? Default (FALSE)
- return_mats
if TRUE, return matrices of the filtered data in a list. Else, return the filtered data.table in bismark format. Default (FALSE)
Value
Either a data.table of filtered data in bismark format with an added
Coverage column (Me + Un) and, if remove_zero_var = TRUE, a Var column with the
per-site variance of Percent; or, if return_mats = TRUE, a list of matrices of
the filtered data with columns in the order of files. Note that
when prop_samples < 1, sites are filtered as a whole: rows from samples below
the coverage threshold are kept at sites that pass.
Details
The function can optionally return a list of CpG x Sample matrices of the filtered data if return_mats = TRUE. The following matrices will be returned for each of the retained CpG sites: Coverage, Percent Methylation (Percent), count of methylated CpGs (Methylated) and count of unmethylated CpGs (Unmethylated)
Examples
# Write three small bismark coverage files
# (chrom, start, end, percent, methylated, unmethylated)
cov <- list(
S1 = "chr1\t100\t100\t50\t5\t5\nchr1\t200\t200\t100\t20\t0\nchr2\t50\t50\t0\t0\t10\n",
S2 = "chr1\t100\t100\t80\t8\t2\nchr1\t200\t200\t100\t30\t0\nchr2\t50\t50\t25\t1\t3\n",
S3 = "chr1\t100\t100\t20\t2\t8\nchr1\t200\t200\t100\t20\t0\nchr2\t50\t50\t40\t4\t6\n"
)
files <- vapply(names(cov), function(n) {
f <- file.path(tempdir(), paste0(n, ".bismark.cov"))
writeLines(cov[[n]], f, sep = "")
f
}, character(1))
# Read in all files using default parameters -- returns a data.table
result <- read_bismark(files)
#>
#> Reading in files...
#> Filtering for Coverage >= 1 in 100% of samples...
#> Done.
# Filter for coverage >= 10 in all samples and remove zero variance features
dt <- read_bismark(files, coverage = 10, remove_zero_var = TRUE)
#>
#> Reading in files...
#> Filtering for Coverage >= 10 in 100% of samples...
#> Removing any zero variance features...
#> Done.
# Same filtering but return a list of CpG x Sample matrices
l <- read_bismark(files, coverage = 10, return_mats = TRUE)
#>
#> Reading in files...
#> Filtering for Coverage >= 10 in 100% of samples...
#> Coercing data to matrices...
#> Done.
l$Coverage
#> S1 S2 S3
#> chr1.100 10 10 10
#> chr1.200 20 30 20
l$Percent
#> S1 S2 S3
#> chr1.100 50 80 20
#> chr1.200 100 100 100
unlink(files)