user3300849 - 1 year ago 93
R Question

# aggregate bins in large GRanges efficiently

I have a SummarizedExperiment (but we can consider it a GRanges).
What I want is to reduce the number of intervals, keeping only one row for every identical adjacent

`mcol(gr)`
, important is to also keep track of the new extend interval.

Thanks a lot!

``````gr <- GRanges(
seqnames = Rle(c("chr1"), c(12)),
ranges = IRanges(1:12*10, end = 1:12*10+5),
state1 = c(1,1,1,1,2,3,4,5,5,5,1,1),
state2 = c(1,1,1,2,2,2,5,5,6,6,1,1))
``````

The resulting GRanges should look like this:

``````gr2 <- GRanges(
seqnames = Rle(c("chr1"), c(8)),
ranges = IRanges(start = c(10,40,50,60,70,80,90,110),
end =  c(35,45,55,65,75,85,105,125)),
state1 = c(1,1,2,3,4,5,5,1), state2 = c(1,2,2,2,5,5,6,1))​
``````

Edit: I have edit the Granges so that a state pair is present also in non adjacent intervals (this second 1,1 pair has to be report independely from the first)
Sorry my initial solution was also wrong!

Thanks a lot!

Create an artificial factor, make sure the levels are in the order of occurrence in the factor (rather than the default alphabetical) to avoid re-arranging the GRanges, and split the `GRanges` object

``````f0 = paste(gr\$state1, gr\$state2, sep=".")
f = factor(f0, levels=unique(f0))
grl = split(gr, f)
``````

Get the ranges and relevant metadata

``````grf = unlist(range(grl), use.names=FALSE)
mcols(grf) = mcols(gr)[!duplicated(f),]
``````

`split()`, `range()`, and `unlist()` should all be 'fast' for data of this size.

To also split on chromosome, add that to the factor

``````f0 = paste(seqnames(gr), gr\$state1, gr\$state2, sep=".")
``````

To split in some other way, e.g., only when states are adjacent, figure out a way to make the appropriate factor, e.g.,

``````f0 = paste(
seqnames(gr),
cumsum(c(TRUE, diff(gr\$state1) != 0)),
cumsum(c(TRUE, diff(gr\$state2) != 0)),
sep=".")
``````