Understanding implications of DP4/Strand Bias in Bacterial VCF files

Question

[f2369583]

Hi,

I'm sorting through some bacterial VCF files generated by bowtie/samtools and a lot of variants have DP4 values indicating the alternative is only on one strand (e.g. DP4 = 5,0,7,0).

The sequences are high-coverage Illumina data.

Is this likely to be sequencing error due to phasing/prephasing rather than a genuine mutation, or could it be strand bias?

Other called variants also yield hits on both strands but do not equal DP (meaning some bases are below the quality threshold); e.g. DP=360, DP4=118,52,54,23 - are these variants likely error?

Thanks!

Answer 1

[Kevin Blighe]

Just looking at the 2 examples that you've mentioned, there looks to be strand bias. The read-depth in the first example is very low, though. There could be various things that could result in strand bias, such as:

• phasing/pre-phasing ( A: Can phasing or pre-phase during basecall cause indel? )
• sequence similarity
• poor quality DNA
• GC content and thermodynamically ultra-fastened (TUF) regions ( https://www.ncbi.nlm.nih.gov/pubmed/22950736 )

You should have an associated P value that indicates strand bias. Also, the information that you've provided should be taken in the context of all other QC information about the data and run before arriving at a decision.