Question

How do I merge a list of GRanges?

2

Entering edit mode

endrebak85 ▴ 40

@endrebak85-10660

Last seen 5.2 years ago

github.com/endrebak/

I have a list of genomicRanges that look like the following:

> l3
[[1]]
GRanges object with 5000 ranges and 4 metadata columns:
         seqnames             ranges strand   |    Island
            <Rle>          <IRanges>  <Rle>   | <integer>
     [1]     chr1     [10050, 10051]      *   |         0
     [2]     chr1     [10100, 10101]      *   |         0
     [3]     chr1     [10200, 10201]      *   |         0
     [4]     chr1     [10250, 10251]      *   |         0
     [5]     chr1     [13250, 13251]      *   |         0
     ...      ...                ...    ... ...       ...
  [4996]     chr1 [1261250, 1261251]      *   |         0
  [4997]     chr1 [1261300, 1261301]      *   |         1
  [4998]     chr1 [1261350, 1261351]      *   |         1
  [4999]     chr1 [1261400, 1261401]      *   |         1
  [5000]     chr1 [1261600, 1261601]      *   |         1
         data.animal.Exp1_12h_PolII.bed data.animal.Exp1_12h_Input.bed
                              <integer>                      <integer>
     [1]                              0                              1
     [2]                              1                              2
     [3]                              1                              1
     [4]                              1                              0
     [5]                              1                              0
     ...                            ...                            ...
  [4996]                              0                              1
  [4997]                              3                              0
  [4998]                              1                              1
  [4999]                              1                              0
  [5000]                              2                              0
         data.animal.Exp2_12h_Input.bed
                              <integer>
     [1]                              1
     [2]                              0
     [3]                              1
     [4]                              0
     [5]                              0
     ...                            ...
  [4996]                              1
  [4997]                              0
  [4998]                              0
  [4999]                              0
  [5000]                              0
  -------
  seqinfo: 1 sequence from an unspecified genome; no seqlengths

[[2]]
GRanges object with 5000 ranges and 5 metadata columns:
         seqnames             ranges strand   |    Island
            <Rle>          <IRanges>  <Rle>   | <integer>
     [1]     chr1     [10000, 10001]      *   |         0
     [2]     chr1     [10050, 10051]      *   |         0
     [3]     chr1     [10100, 10101]      *   |         0
     [4]     chr1     [10150, 10151]      *   |         0
     [5]     chr1     [10200, 10201]      *   |         0
     ...      ...                ...    ... ...       ...
  [4996]     chr1 [1119900, 1119901]      *   |         0
  [4997]     chr1 [1120000, 1120001]      *   |         0
  [4998]     chr1 [1120100, 1120101]      *   |         0
  [4999]     chr1 [1120150, 1120151]      *   |         0
  [5000]     chr1 [1120200, 1120201]      *   |         0
         data.animal.Exp1_15h_PolII.bed data.animal.Exp2_15h_PolII.bed
                              <integer>                      <integer>
     [1]                              1                              0
     [2]                              0                              0
     [3]                              2                              0
     [4]                              1                              0
     [5]                              0                              0
     ...                            ...                            ...
  [4996]                              0                              0
  [4997]                              0                              0
  [4998]                              0                              0
  [4999]                              0                              0
  [5000]                              0                              0
         data.animal.Exp1_15h_Input.bed data.animal.Exp2_15h_Input.bed
                              <integer>                      <integer>
     [1]                              0                              2
     [2]                              1                              0
     [3]                              0                              1
     [4]                              0                              0
     [5]                              1                              1
     ...                            ...                            ...
  [4996]                              0                              1
  [4997]                              0                              1
  [4998]                              1                              0
  [4999]                              1                              0
  [5000]                              0                              1
  -------
  seqinfo: 1 sequence from an unspecified genome; no seqlengths

The real list will contain many more granges. Notice that there are several data columns in each file.

What I want is a GRanges object containing the union of all the ranges and the metadata. I could loop over the list, taking the union of each pair of GRanges, but the union loses the metadata.

Is there a way to merge all my GRanges into one giant GRanges, with metadata? (What I ultimately want is a dataframe of counts so that I can input it into limma.)

GenomicRanges • 20k views

ADD COMMENT • link updated 2.8 years ago by jon.klonowski ▴ 40 • written 8.4 years ago by endrebak85 ▴ 40

1

Entering edit mode

I am curious as to how you get this list of GRanges. Wouldn't it make sense to create an eSet or SummarizedExperiment for use with limma?

ADD REPLY • link 8.4 years ago Michael Lawrence ★ 11k

0

Entering edit mode

Ah, so if I have multiple granges (with different, somewhat overlapping rows) in a summarizedexperiment, limma can treat it like one matrix? Voom even? (I have very little bioconductor experience, unfortunately.)

ADD REPLY • link 8.4 years ago endrebak85 ▴ 40

6

Entering edit mode

merv ▴ 140

@mmfansler-13248

Last seen 21 days ago

MSKCC | New York, NY

I suspect this may have changed, but with Bioconductor 3.12 packages, I can do

unlist(as(l3, "GRangesList"))

which converts the base list object to a GRangesList and then uses the GRL-specific unlist to merge to single GRanges object.

ADD COMMENT • link 4.0 years ago • updated 3.8 years ago merv ▴ 140

1

Entering edit mode

This was the only solution that worked for me, thanks! But don't forget closing parentheses :P

ADD REPLY • link 3.8 years ago James ▴ 10

score 6 · Accepted Answer · 2016-06-21

6

Entering edit mode

Hervé Pagès 16k

@herve-pages-1542

Last seen 20 hours ago

Seattle, WA, United States

Hi,

Merge is a pretty vague term. My understanding is that you want to concatenate all the GRanges objects in the list. In the R world, this concatenation is performed with c().One complication here is that the arguments to your call to c() are in a list. So you need to use do.call("c", list_of_arguments):

l1 <- list(1:5, 21:22, 31:34)
do.call("c", l1)  # same as c(l1[[1]], l1[[2]], l1[[3]])
# [1]  1  2  3  4  5 21 22 31 32 33 34

So with your list of GRanges objects do.call("c", l3) should do what you want.

H.

ADD COMMENT • link 8.4 years ago Hervé Pagès 16k

0

Entering edit mode

Thanks. What I want to do is create a new granges object. If it is a concat, it is a horizontal concat. The new granges should have n * 4 metadata columns, where n is the number of granges, 4 is the number of metadata columns in each original grange.

So if I have two granges

a:

chr1 100 101 5 6

chr 200 201 7 8

and b:

chr 150 151 0 1

chr 200 201 1 1

I want to create a new granges object like this (a + b):

chr1 100 101 5 6 0 0

chr  150 151 0 0 0 1

chr 200 201 7 8 1 1

This should preferably work for an arbitrary number of granges objects. Perhaps this is not possible? Sorry for being unclear in my original q.

ADD REPLY • link 8.4 years ago endrebak85 ▴ 40

1

Entering edit mode

I am working on a merge() method now. It is only a binary merge but it should do the trick with Reduce().

ADD REPLY • link 8.4 years ago Michael Lawrence ★ 11k

3

Entering edit mode

Ok, added in S4Vectors 0.11.7, Reduce(merge, l3) should work. NAs are filled into the missing cells, so if you want those as zeros, you will need to convert them explicitly.

ADD REPLY • link 8.4 years ago Michael Lawrence ★ 11k

0

Entering edit mode

do.call method did not work for me (the result remained a list), however, the following from "GRangesList-class {GenomicRanges}" R documentation did:

"unlist(x, recursive = TRUE, use.names = TRUE): Concatenates the elements of x into a single GRanges object."

I'm still looking for a method to produce a union of all lines in the resulting GRanges object (so I end up with fewer but larger intervals), preferably without looping.

ADD REPLY • link 7.4 years ago lovro0 ▴ 10

0

Entering edit mode

Concatenating two lists results in a list, so I guess you're seeing the expected behavior. Probably you want to provide a reproducible example of the data you are starting with (minimal, just a few lines to create an artificial data object...) and some explicit indication of what you want at the end.

?"relist,ANY,PartitioningByEnd-method" mentions relist() (typically used as unlist, transform, relist for 'vectorized' operations on ranges rather than iterative operation on elements of a list) and regroup() (form into new skeleton without intermediate operation). 'union' might be the within-GRanges operation 'reduce()' ?

ADD REPLY • link 7.4 years ago Martin Morgan 25k