bedtools intersect with "bedgraph" style with multiple extra columns of data for multiple samples?

Question

[dansapozhnikov]

Hi,

I can successfully perform bedtools intersect for two 3-column BED files, but I want to perform the same intersect using a "file A" that contains multiple additional columns with values for a large number of samples formatted like so:

File A:

(chr) (start) (stop) (sample 1..2..3...4...etc..)

chr1   1001   1002    10    25   14   25
chr1   1006   1007    12    22   11   50
chr1   1012   1013    44    11   12   30

File B:

chr1   500    1010
chr1   2000   3000

Output:

chr1   1001   1002    10    25   14   25
chr1   1006   1007    12    22   11   50

Does anyone know if there is a way to use bedtools intersect, or bedtools intersect with another program, or any other code at all into perform this kind of intersect and retain the data in the additional columns in the output?

Thanks!

Dan

[ATpoint]

Have a look at bedtools unionbedg.

[dansapozhnikov]

Actually, the File A is an output of unionbedg. So it is a bedgraph with many columns and then I want to intersect it with a BED file with intervals. (Also File A has data for each single nucleotide whereas the BED file has intervals)

I added an example to the original post.

Answer 1

[Alex Reynolds]

Not sure if/how your files are sorted, but this should take care of that:

$ bedops -e 1 <(sort-bed fileA.bed) <(sort-bed fileB.bed) > answer.bed

[dansapozhnikov]

This does not work for me for some reason. It says "Non-numeric end coordinate. See line 1 in fileB.bed". This is not even the problematic file and it worked in bedtools intersect.

[Alex Reynolds]

If you run cat -te on a few lines of your files, what does it say? For instance what comes out of: head -5 foo.bed | cat -te? Something's up with your files, which needs fixing.

[dansapozhnikov]

chr1^I2572970^I2579715^M$

[Alex Reynolds]

See that ^M at the end? That's a Windows carriage return character. You need to remove that. You can use tr for this:

$ tr -d '\r' < foo.bed > foo.fixed.bed

Repeat for all afflicted files, then run your commands on the fixed files.

[dansapozhnikov]

OK so this seems to be a Windows line ending, so I ran dos2unix. Then I ran bedops as before, and now (unlike bedtools intersect) it ran, but it does not retain the extra columns.

[Alex Reynolds]

What is the output of this:

$ head -5 fileA.fixed.bed | cat -te

What is the output of this:

$ head -5 fileB.fixed.bed | cat -te

Please post everything you see. The bedops -e command just reports back any elements as they are found, and does not modify them. So either your files are not structured as described, or there is some other problem. If we can see your actual inputs, we can probably figure out what's up.

[dansapozhnikov]

Thanks to your suggestions I found the problem! File A was space-delimited, not tab. It now runs and retains all the columns. Thank you very much for your help.

[Alex Reynolds]

Awesome! Glad to help.