`.make_bed()` can write negative BED start coordinates near chromosome starts

[Gouhuan1999]

.make_bed() builds a windowed BED around a focal point but never clamps the lower bound to 0, so features close to the start of a chromosome get negative coordinates.

R/utils_data.R:43:

  if (!for_profile) {
    if (tss == "center") {
      bed[, focal_point := as.integer(apply(bed[,2:3], 1, mean))]
      bed[, bed_start := focal_point-up]
      bed[, bed_end := focal_point+down]
    } else if (tss == "start") {
      bed[, bed_start := start-up]
      bed[, bed_end := start+down]
    } else {
      bed[, bed_start := end-up]
      bed[, bed_end := end+down]
    }
    bed <- bed[, .(chr, bed_start, bed_end)]
    data.table::setkey(x = bed, chr, bed_start, bed_end)
  }

With the default up = 2500, any focal point < 2500 bp from the chromosome start yields bed_start < 0. The file is then written out verbatim by data.table::fwrite(...). Negative starts are invalid BED and downstream tools (bwtool / bedtools, which the header comment references) will error or silently skip those rows, so a handful of near-telomeric/near-start features quietly drop out of profile/aggregation plots.

Two smaller things in the same function worth folding in:

• as.integer(apply(bed[,2:3], 1, mean)) truncates toward zero rather than rounding, so the center is off by up to 1 bp versus round().
• The for_profile = TRUE path skips the bed[, .(chr, bed_start, bed_end)] projection entirely and writes whatever extra columns bed still carries, even though the comment on the first line states "bwtool tool requires only three columns."

Suggested fix: clamp with pmax(0L, ...) on the start (BED is 0-based, so 0 is the floor), e.g. bed[, bed_start := pmax(0L, focal_point - up)], and mirror it in the start/end branches.