Bedtools merge but only if intersection exists

Question

[jrose]

I want to instead of use the -wa option or -wb option in bedtools intersect, i want the union of all regions which have some intersection.

So if my input is like so (first row bed 1 second row bed 2):

ccccc        ccccc   cc   ccccccccc
   cccc                 ccccccccccccccccccc

then I want the output to be:

ccccccc                 ccccccccccccccccccc

Is there an operation like this?

I can do bedtools intersect -wa and then -wb on the same input and now I have to files only of regions that contain an intersection, and then take their union, but i was wondering if there was something already implemented.

[ATpoint]

It's as you say just a pipe of two commands, I do not see the downside of that:

bedtools intersect -a first.bed -b second.bed -wa -wb | bedtools merge -i -

See answer.

[jrose]

ah ok. I did not know you could do -wa -wb. Thanks

[ATpoint]

requires some awk, see below

[jrose]

Thank you all

Answer 1

[ATpoint]

Assuming a three-column BED file you could do this, with an awk command as -wa -wb will make a 6-column BED file after the intersect:

% cat first.bed 
chr1    1   5
chr1    10  20
chr1    30  40

% cat second.bed 
chr1    4   8
chr1    15  18
chr1    100 200

% bedtools intersect -a first.bed -b second.bed -wa -wb                                                    
chr1    1   5   chr1    4   8
chr1    10  20  chr1    15  18

% bedtools intersect -a first.bed -b second.bed -wa -wb | awk 'OFS="\t" {print $1, $2, $3"\n"$4, $5, $6}'                                                 
chr1    1   5
chr1    4   8
chr1    10  20
chr1    15  18

# and finally all in one plus merge:
% bedtools intersect -a first.bed -b second.bed -wa -wb | awk 'OFS="\t" {print $1, $2, $3"\n"$4, $5, $6}' | sort -k1,1 -k2,2n | bedtools merge -i -
chr1    1   8
chr1    10  20

BEDtools version 2.30.0

[jrose]

Actually my bedtools v2.27.1 returns a 3 column bed file with -wa -wb

[ATpoint]

I am on 2.30.0.

Answer 2

[Alex Reynolds]

Get all elements of first that have one or more overlaps with elements in second, and vice versa, and then merge those two sets:

% bedops --merge <(bedops -e 1 first.bed second.bed) <(bedops -e 1 second.bed first.bed)
chr1    1   8
chr1    10  20

Each of these is a process substitution that calculates overlaps:

<(bedops -e 1 first.bed second.bed)

and:

<(bedops -e 1 second.bed first.bed)

You can use process substitutions in place of files with BEDOPS tools.