Handling get_variant_context function's "both alleles have consequences" warning

[youngchanpark]

There's this part of code in the get_variant_context function.

man_qq_annotate/manqq_functions.R

Line 372 in aa79408

for(i in alleles) {

for(i in alleles) {
          #print(paste("allele", i))
	       #print(colnames(cons))
	       topaste=getVepSnp(chr=chr,pos=pos,allele=i,build=build)
	       #print(colnames(topaste))
	       #if(!("colocated_variants" %in% colnames(topaste))){topaste$colocated_variants=""}

	        #cons=ifelse(is.null(cons), topaste, rbind(cons,topaste))
          if(!is.null(topaste)){
            if(!is.null(cons)){warning("problem: both alleles have consequences");print(cons); print(tobind)}
            cons=topaste$most_severe_consequence[1]
          }
      }

There is currently a code that outputs a warning message that says both alleles have a consequence assigned to it.

I found a real situation where this occurred:

chr = 17 
pos = 5762108
a1 = 'CGT'
a2 = 'CGTGT'

a1_vep = getVepSnp(chr=chr,pos=pos,allele=a1,build=build)
a2_vep = getVepSnp(chr=chr,pos=pos,allele=a2,build=build)

print(a1_vep$most_severe_consequence)
print(a2_vep$most_severe_consequence)

Both allele's consequence is intergenic, but I thought I'd assume there would be another case where the consequence may differ.

The current code just prints out the a1's consequence and overwrites the cons variable with a2's consequence value.

I'm writing up this issue just to make sure whether this is our intended behaviour, or whether we need to handle it differently.

If this is indeed an intended behaviour, I think we should add a line of warning message, saying the A2's consequence will overwrite the consequence.

Would like your input @agilly

[agilly]

I presume this will happen mostly for indels where the REF is not exactly the REF that Ensembl has, or for multiallelics.

How is the most_severe_consequence field handled down the line? Is it parsed and conditioned on in any meaningful way? If not, ideal behaviour could be to make it something like "CGT=CGTGT=intergenic" if in cases like these, the csq are equal, or "AT=synonymous,ATAT=frameshift" if they are not.