Need help with BCFtools annotate; not all information is carried forward

Question

[Hadi M]

Hi everyone,

I have a custom tab delimited annotation file that I used to annotate a VCF file using bcftools annotate. It worked just fine but the only problem is that if there are multiple hits on a particular position, only the first information is carried into the VCF file. Here's an example:

This is an example of my annotation file:

#CHROM  FROM    TO  TRAIT
chr3    100001  100010  Disease A
chr3    100005  100005  Disease B
chr3    100005  100005  Disease C

And here's an example of my VCF file:

#CHROM  POS ID  REF ALT
chr3    100005  .   A   T

Annotating the VCF file produce:

#CHROM  POS ID  REF ALT INFO
chr3    100005  .   A   T   Disease A

As you can see, only the first information is carried forward. My ideal output is:

#CHROM  POS ID  REF ALT INFO
chr3    100005  .   A   T   Disease A
chr3    100005  .   A   T   Disease B
chr3    100005  .   A   T   Disease C

Or:

 #CHROM POS ID  REF ALT INFO
 chr3   100005  .   A   T   Disease A | Disease B | Disease C

Is there an option in bcftools annotate that would allow me to get such output? If there is an alternative tool, do recommend as well. Cheers.

Answer 1

[Ram]

It is not legal for a VCF file to have multiple entries for the same chr-pos-ref-alt combination.

Option 1:

Try the --merge-logic parameter in bcftools annotate. I've never tried it, but it looks like it might work when used in the manner --merge-logic TRAIT:unique

Option 2:

You should be able to use R (dplyr) to get a new annotation file from your existing annotation file. Group by CHROM, POS, ID, REF, ALT and aggregate TRAIT to paste(TRAIT, collapse = " | "). This, of course, has the downside that range annotations and point annotations cannot be aggregated together (the group-by will only group by identical values, not overlapping ranges), forcing you to convert all range annotations to point annotations before aggregating them.