Overlapping Ranges within Granges object

Question

[ntsopoul]

Hi,

I have designed a Crispr library with dual guide RNAs flanking specific genomic regions. I used crisprDesign to do this.

I have now too many gRNA pairs targeting the same region. I have the data as GRanges objects.

Is there a way to filter the IRanges in a Granges object for non-overlapping regions or at least allow 10% overlap? I tried findOverlaps but with no avail.

Answer 1

[Alex Reynolds]

If you're not tied to Granges, you could use bedmap --fraction-both 0.1 to require at least 10% overlap between reference and map regions. Then use the bedops --not-element-of operation to get all elements not in this overlap set (i.e., regions not overlapping or having up to 9.99999% overlap).

[ntsopoul]

does this work with a single .bed file?

[Alex Reynolds]

Yes, bedmap will perform operations on one or two BED files.

Answer 2

[ATpoint]

There is no one-hit function in GenomicRanges, but you can stick something together using a combination of findOverlaps to first find overlaps and then pintersect + width to count the bases that overlap. THen you just need to do a calculation to put the overlap with in relation to some reference width to get a percentage:

a <- GRanges(seqnames = "chr1", ranges = IRanges(start=c(1, 20), end=c(10, 200)))
b <- GRanges(seqnames = "chr1", ranges = IRanges(start=c(5, 500), end=c(10, 1000)))

fo <- findOverlaps(a, b)
> fo
Hits object with 1 hit and 0 metadata columns:
      queryHits subjectHits
      <integer>   <integer>
  [1]         1           1
  -------
  queryLength: 2 / subjectLength: 2

wi <- width(pintersect(a[from(fo)], b[to(fo)]))
6

[ntsopoul]

I have one Granges object and not two. Every IRanges entry is the critical region of the gRNA. Can this still work?

[ATpoint]

I would probably collapse overlapping regions with reduce to create the second GRanges. Can you add some representative data? Using dput?

[ntsopoul]

Here is some representative data and an image in igv of one gene. I have loaded the .bed file in igv. As you can see I have many regions that are overlapping, however, I just need one gRNA pair to cover a stratech of the genomic region. I would like to filter the bed file for one of this blocks and remove the rest that overlap. Ideally, I would like to have some minor overlap but this is not so important at the moment.

GRanges object with 1467 ranges and 0 metadata columns:
         seqnames            ranges strand
            <Rle>         <IRanges>  <Rle>
     [1]     chr2 90735572-90736199      -
     [2]     chr2 90735560-90736199      +
     [3]     chr2 90735621-90736199      -
     [4]     chr2 90735659-90736260      +
     [5]     chr2 90735572-90736243      -
     ...      ...               ...    ...
  [1463]     chr2 90738687-90739243      -
  [1464]     chr2 90738669-90739243      +
  [1465]     chr2 90738692-90739243      +
  [1466]     chr2 90738693-90739243      +
  [1467]     chr2 90738728-90739243      -
  -------
  seqinfo: 1 sequence from an unspecified genome; no seqlengths

[ntsopoul]

you are on to something… if I use reduce to find the “overlaping regions” in a secon GRanges object, what do I do next?