Persistent zero SNPs WARNING

[paul-manchester]

Hi

I have seen this issue raised before but there doesnt seem to be an answer - please could you have a look at this?

Im getting a lot of

WARNING: In file temp_vcf4_C169.vcf, in AFUA_1G09640, near site 2497803,

there is a stretch of ≥1000 codons with zero SNPs called. If this was unexpected,

you may need to specify Unix (\n) newline characters! See Troubleshooting.

I followed the troubleshooting section and checked the line characters - no issues and its all handled in UNIX
There are indeed many SNPs in these regions.
for example (different region but the error is the same across the entire vcf) - >99% zero results with occasional hits such as the AFUA_1G00550 below.

#output in product_results file:

file product N_diffs S_diffs N_diffs_vs_ref S_diffs_vs_ref N_sites S_sites piN piS mean_dN_vs_ref mean_dS_vs_ref mean_gdiv_polymorphic mean_N_gdiv mean_S_gdiv
temp_vcf4_C121.vcf AFUA_1G00510 0 0 0 0 1315.66666666667 421.333333333332 0 0 0 0 * * *
temp_vcf4_C121.vcf AFUA_1G00520 0 0 0 0 1849.33333333334 556.666666666666 0 0 0 0 * * *
temp_vcf4_C121.vcf AFUA_1G00540 0 0 0 0 1832.5 594.5 0 0 0 0 * * *
temp_vcf4_C121.vcf AFUA_1G00550 0.403684207330771 0.0909090909090908 3.88078541374474 1.95454545454545 1076.25494071146 337.017786561265 0.000375082326743062 0.000269745676739126 0.00360582355253054 0.00579953204989123 0.0676130356306299 0.128838129970092 0.0867768595041321

[paul-manchester]

sorry - here is some of the data for this region
https://www.dropbox.com/scl/fi/tvj34be7zjbadpgk20jtj/test-data?rlkey=3nivns36n2l8qlv77lys1j63k&st=ue6o8wyb&dl=0

[singing-scientist]

Hi @paul-manchester! Thanks for reaching out about this. The key words in this warning are "if this was unexpected"—it's not necessarily a bug or problem, just something to be aware of. SNPGenie was originally written for high-diversity viral samples where going >1000 codons without seeing a single variant would, in fact, be unexpected. So the real question is, given your SNP input, would this be unexpected? As long as you can verify that variants are being recorded at the right sites/codons, this is OK.

[paul-manchester]

Thanks for clarifying and getting back so quickly - its a set of fungal genomes so, yes you are exactly right - there are sometimes stretches like this - I commented out the detection/warning lines to stop the log getting too big
The next issue was a problem with multi-allelic variants which are throwing errors - looks like merging vcfs into a large GVCF format fils is not correctly porting the AD values which then throws the script off. Is there a way to allow poor AD calls through? It works fine if I just convert to biallelic calls so its only a niggle really...

[singing-scientist]

My pleasure! I was shall we say a less-than-polished coder back then, so apologies for the way errors can be redundant, nonideal.

For multi-allelic positions, these are always such a hassle. SNPGenie does handle them correctly, but is very finicky about input. It’s an unfortunate state of affairs that there are essentially as many VCF formats as there as VCF callers/settings. So I took the approach of just requiring the user to figure out how to get their input to comply with one of the few input types I allow. One approach is just to format everything in the VCF type 4 or CLC style, and everything will get handled. Another approach might be to just filter to biallelic and making note where variants were eliminated, checking those positions later. I’m sorry there’s no “right” answer, but ultimately designing SNPGenie to accept every possible input was an insurmountable task, and I leave the input format problem to the user.

I was unable to see what file type the Dropbox link was, as there was no extension. Feel free to share a VCF sample or a few lines here if you’d like more feedback.

Cheers,
Chase