Overlapping variants in fail.vcf and pass.vcf cause preconsensus to be masked and thus bcftools complains when applying the "pass" variant

[Sam-Sims]

Hello,

I think there are edge cases that result in overlapping variants in both the pass and fail vcf files causing bcftools to complain and artic minion to fail:

My fail.vcf contains a variant at 21640:

MN908947.3 21640 . TG GA 65 PASS DP=43;AC=0,2;AM=41;MC=0;MF=0.0;MB=0.0;AQ=3.65;GM=1;PH=6.02,6.02,6.02,6.02;SC=None; GT:GQ:DP:PS:UG:UQ 0/1:32:43:.:0/1:31.97

Meanwhile the pass.vcf contains a 9nt deletion at 21632, overlapping 21640:

MN908947.3 21632 . TTACCCCCTG T 500 PASS DP=43;AC=0,43;AM=0;MC=0;MF=0.0;MB=0.0;AQ=28.59;GM=1;PH=6.02,6.02,6.02,6.02;SC=None; GT:GQ:DP:PS:UG:UQ 1/1:137:43:.:1/1:136.04

Due to the variant being present in the fail.vcf this is maksed in the preconsensus when artic_mask is run and so TG becomes NN.

When bcftools is run it expects to find TTACCCCCTG since that is the variant in the pass.vcf file - but instead is finding the masked TTACCCCCNN in the preconsensus and so complains:

The fasta sequence does not match the REF allele at MN908947.3:21632:    
   REF .vcf: [TTACCCCCTG]  
   ALT .vcf: [T]
   REF .fa : [TTACCCCCNN]

As a temp workaround I have masked the variant in the pass.vcf and re-run bcftools manually to generate a consensus sequence. Do you have any reccomendations for these edge cases?

[Sam-Sims]

Looking into this a bit further it appears that in this sample amplicon 72 has dropped - which is in pool 2.

Looking at the *.2.vcf file generated from medaka variant ./primer-schemes/scov2/V4.1/scov2.reference.fasta samplename.2.hdf samplename.2.vcf it contains 2 variants from the dropped amplicon 72 that overlap with the deletion found in *.1.vcf:

MN908947.3  21634   .   ACCCC   GTGG    9.798   PASS    .   GT:GQ   1:10
MN908947.3  21640   .   TG  GA  6.765   PASS    .   GT:GQ   1:7

These variants are not found in the *.1.vcf so I assume are just artifacts coming from the dropped amplicon.

Interestingly when running with the --no-longshot flag and instead use medaka tools annotate you can see the depth of each variant is 1 in the *.2.vcf which is to be expected since they came from the dropped amplicon. They are then filtered out correctly and minion completes with no errors.

MN908947.3  21634   .   ACCCC   GTGG    9.798   PASS    DP=1.0;DPS=1.0,0.0  GT:GQ   1:10
MN908947.3  21640   .   TG  GA  6.765   PASS    DP=1.0;DPS=1.0,0.0  GT:GQ   1:7

However if longshot is used it seems like the variant at 21640 is left in and so the merged.vcf contains the following entry with a depth of 43 .

MN908947.3  21640   .   TG  GA  65  PASS    DP=43;AC=0,2;AM=41;MC=0;MF=0.000;MB=0.000;AQ=3.65;GM=1;PH=6.02,6.02,6.02,6.02;SC=None;  GT:GQ:DP:PS:UG:UQ   0/1:32:43:.:0/1:31.97

This is written to the fail.vcf still - I am guessing because of the quality score (?) and then causes the issue detailed above.

Perhaps this is a bug with longshot when calculating the depth of the variant as I am unsure where the depth of 43 has come from - nothing in the BAM indicates why this would be.

Running minion with --strict works too, but the deletion is omitted from the consensus.

Admittedly this does seem to be an edge case resulting from a dropped amplicon - which isnt ideal in the first place.

[knewl]

Hello,

I am getting the above error in ARTIC v1.2.4 in bcftools consensus due to overlapping 'pass' and 'fail' variants.

 Running: bcftools consensus -f sample_name.preconsensus.fasta sample_name.pass.vcf.gz -m sample_name.coverage_mask.txt -o sample_name.consensus.fasta

  The fasta sequence does not match the REF allele at MN908947.3:21632:
     REF .vcf: [TTACCCCCTG]
     ALT .vcf: [T]
     REF .fa : [NNNNNNNNNN]CATAC

This is caused by an 11bp deletion in fail.vcf overlapping with a 10bp deletion in pass.vcf. The 11bp deletion is added to the mask and thus the preconsensus contains Ns which overlap with the variant which passes.

fail.vcf:

MN908947.3      21631   .       ATTACCCCCTG     A       309.485 PASS    DP=13.0;DPS=5.0,8.0;Pool=2      GT:GQ   1:309

pass.vcf:

MN908947.3      21632   .       TTACCCCCTG      T       289.132 PASS    DP=45.0;DPS=13.0,32.0;Pool=1    GT:GQ   1:289

This is a very similar issue to #21 which has been fixed . The difference is that in this case the variants do not have exactly the same starting position, they simply overlap. A solution would be to ignore failed variants which overlap with a variant which passes.