Sliding window methylation summary along the genome

Computes a per-position sliding window statistic (median or mean) of methylation beta values for each sample in a commaData object. The genome size for each chromosome is read from genome(object), so no organism-specific values are ever hardcoded. Circular genome wrap-around is supported via .circularIndex().

Usage

slidingWindow(
  object,
  window,
  stat = c("median", "mean"),
  mod_type = NULL,
  circular = TRUE
)

Arguments

object: A commaData object. Must have genome size information in genome(object) (i.e., it must have been constructed with a genome argument).
window: Positive integer. Window size in base pairs. The smoothed value at position \(p\) is computed from positions \([p - \lfloor w/2 \rfloor,\; p + \lfloor w/2 \rfloor]\).
stat: Character string. Summary statistic to apply within each window. One of "median" (default) or "mean".
mod_type: Character string or NULL. If provided, only sites of the specified modification type (e.g., "6mA") are included in the smoothing. If NULL (default), all sites are used.
circular: Logical. If TRUE (default), positions at the ends of each chromosome are wrapped around so that the window at position 1 can draw from positions near the chromosome end, and vice versa. Appropriate for circular bacterial chromosomes.

Value

A data.frame with one row per (chromosome, position, sample) combination, containing:

chrom: Chromosome name (character).
position: Genomic position, 1-based (integer).
sample_name: Sample identifier (character).
window_stat: Smoothed beta value (numeric). The column is named window_median or window_mean depending on stat.

Details

Because most positions in a genome have no methylation site, the beta value vector for each chromosome-sample pair is sparse (mostly NA). zoo::rollapply is called with na.rm = TRUE so that windows spanning regions with no data still produce a value where at least one site is present.

Positions where every site in the window is NA (i.e., no coverage) remain NA in the output.

Examples

data(comma_example_data)
# \donttest{
sw <- slidingWindow(comma_example_data, window = 5000L)
head(sw)
#>     chrom position sample_name window_median
#> 1 chr_sim        1      ctrl_1     0.8581977
#> 2 chr_sim        2      ctrl_1     0.8581977
#> 3 chr_sim        3      ctrl_1     0.8581977
#> 4 chr_sim        4      ctrl_1     0.8581977
#> 5 chr_sim        5      ctrl_1     0.8581977
#> 6 chr_sim        6      ctrl_1     0.8581977
# Filter to one sample
sw_ctrl1 <- sw[sw$sample_name == "ctrl_1", ]
# }