Issue with ExAC and 1000g hg38 lifted-over data and systematic failure of annotation softwares!

Question

[reza.jabal]

ExAC data with hg38 coordinates has been around for filter-based annotation since late 2015, but it seems there is a systematic problem with the use of ExAC and 1000G lifted-over data data for annotation! Mainstream annotation softwares (Annovar, VEP and snpEff) fail to incorporate MAF for variants that their corresponding contig is reversed in the hg38 assembly. As a result, common variants in ExAC and 1000G populations might be misinterpreted as novel variant solely because annotation softwares fail to report corresponding MAF.

I was wondering if anyone here has come across the same problem and if so, how they have tackled this problem?

Answer 1

[Pablo]

In my opinion, this sounds like a problem in the lift-over procedure, as opposed to a failure in the annotation software.

In order to correctly lift-over variants (e.g. a VCF file from ExAC), not only the coordinates should be changed, but also the variant's REF and ALT fields must be complemented accordingly in reversed genes (I'm talking about WC-complement). If this last part is not done right, downstream annotation software would fail to annotate just because the input is incorrect.

Again, this is an opinion / guess (since there was no sample data in the post, I cannot dig deeper).

[reza.jabal]

Hi Pablo, Thanks for your comment! Yes you guessed it right. Deeper investigation of the matter led me to realisation that the problem rather lays in dbsnp liftedover data. I am now using a custom script to fill in missing frequencies.