bedtool closest

Question

[CHINMAYA]

If I have a reference BED file and second file containg ncRNA BED file.

I want to compute the distance of ncRNAs to the nearest protein-coding genes and remove gene in ncRNA located within 250 bp from protein-coding genes on the same strand.

Can anyone help?

[Ram]

Side note: It's bedtools, not "bedtool".

Answer 1

[Alex Reynolds]

If you can use the bedops kit, you can use closest-features to solve this:

1. Separate reference and ncRNA files by strand and sort:

   $ awk -v FS="\t" -v OFS="\t" '($6 == "+")' reference.bed | sort-bed - > reference.for.bed
   $ awk -v FS="\t" -v OFS="\t" '($6 == "-")' reference.bed | sort-bed - > reference.rev.bed
   $ awk -v FS="\t" -v OFS="\t" '($6 == "+")' ncRNA.bed | sort-bed - > ncRNA.for.bed
   $ awk -v FS="\t" -v OFS="\t" '($6 == "-")' ncRNA.bed | sort-bed - > ncRNA.rev.bed
   

2. Filter by positive distances for forward-strand elements. To confirm from your question, you want to remove elements that are within 250nt of the reference element's TSS (start position), so the distance reported from --dist --closest will be negative:

   $ closest-features --dist --closest reference.for.bed ncRNA.for.bed | awk -v FS="|" -v OFS="\t" '($3 < -250)' > filtered.for.bed
   

3. Repeat for reverse-stranded elements. In this case, the threshold will be positive, because the TSS of reverse-stranded elements will be at the stop position:

   $ closest-features --dist --closest reference.rev.bed ncRNA.rev.bed | awk -v FS="|" -v OFS="\t" '($3 > 250)' > filtered.rev.bed
   

4. Take the union of the filtered results:

   $ bedops --everything filtered.for.bed filtered.rev.bed > filtered.bed
   

References:

• https://bedops.readthedocs.io/en/latest/content/reference/set-operations/closest-features.html
• https://bedops.readthedocs.io/en/latest/content/reference/set-operations/bedops.html#everything-u-everything