Bacterial genomes and snpeff warnings 'WARNING_TRANSCRIPT_NO_START_CODON'

Question

I annotated a bunch of genomes with snpeff. And although the original reads' qualities and the matches with the reference genomes were quite high, I see the above mentioned warning in almost every variant in my vcf file. Thus, I wonder how this could happen. Since I'm using bacterial genomes with a manually built database (from these reference genomes) and there are quite some genes (especially in gene clusters) that have the alternative start codon GTG, I suspect this causing the warning. Any idea how to test if this is really the reason for the warning?

Thanks a lot in advance!

Edit: As requested, here is the html snpEff report. And I've also uploaded an example vcf file after annotation here

Answer 1

Bacteria have a different translation table from eukaryote nuclear DNA; it may be that SnpEff is expecting human genes in the genome. Have you followed the build steps outlined in the documentation?

https://pcingola.github.io/SnpEff/se_build_db/

If your genome, or a chromosome, uses non-standard codon tables you must update snpEff.config accordingly. A typical case is when you use mitochondrial DNA. Then you specify that chromosome 'MT' uses codon. Invertebrate_Mitochondrial codon table. Another common case is when you are adding a bacterial genome, then you specify that the codon table is Bacterial_and_Plant_Plastid.

Answer 2

According to your report, you are using human HG18 as the reference genome version. Therefore, the report will be wrong along with the wrong codons warnings. It is unfruitful to try with a human database with bacteria.