Reducing and aggregating GRanges with gaps using plyranges.

Question

[rpolicastro]

I just started using plyranges, and I can not figure out how to reduce and aggregate a GRanges object with a desired gap width.

Example data.

library("plyranges")

df <- data.frame(
  seqnames="chrI", start=c(1, 10, 20), end=c(5, 15, 25), strand=c("+", "+", "-"),
  score=c(8, 3, 6)
)
gr <- as_granges(df)

> gr
GRanges object with 3 ranges and 1 metadata column:
      seqnames    ranges strand |     score
         <Rle> <IRanges>  <Rle> | <integer>
  [1]     chrI       1-5      + |         8
  [2]     chrI     10-15      + |         3
  [3]     chrI     20-25      - |         6
  -------
  seqinfo: 1 sequence from an unspecified genome; no seqlengths

Desired output with max allowed gap width of 10 and summing the scores for the aggregation in this example.

desired_output <- data.frame(
  seqnames="chrI", start=c(1, 20), end=c(15, 25), strand=c("+", "-"),
  score=c(11, 6)
)
desired_output <- as_granges(desired_output)

> desired_output
GRanges object with 2 ranges and 1 metadata column:
      seqnames    ranges strand |     score
         <Rle> <IRanges>  <Rle> | <numeric>
  [1]     chrI      1-15      + |        11
  [2]     chrI     20-25      - |         6
  -------
  seqinfo: 1 sequence from an unspecified genome; no seqlengths

This is similar to section 4.1 in the HelloRanges tutorial, which does work for me since you can set a minimum gap width in the GenomicRanges::reduce function. The plyranges equivalent is reduce_ranges_directed but it does not appear to have a gap width option.

EDIT: This has been cross-posted to bioconductor support also.

[Friederike]

Will this do?

stretch(anchor_start(gr), extend=10) %>% reduce_ranges_directed(., sum.score = sum(score))
GRanges object with 2 ranges and 1 metadata column:
      seqnames    ranges strand | sum.score
         <Rle> <IRanges>  <Rle> | <numeric>
  [1]     chrI      1-25      + |        11
  [2]     chrI     20-35      - |         6
  -------
  seqinfo: 1 sequence from an unspecified genome; no seqlengths

EDIT: replaced flank_right with stretch.

But I can see how it might be annoying that the result will have aberrant bp added to the final interval