The main challenge associated with non-diploid variant calling is the difficulty in distinguishing between the sequencing noise (abundant in all NGS platforms) and true low frequency variants. Some of the early attempts to do this well have been accomplished on human mitochondrial DNA although the same approaches will work equally good on viral and bacterial genomes (Rebolledo-Jaramillo et al. 2014, Li et al. 2015).
Comments
FreeBayes is widely used for calling variants in diploid systems. However, it can also be used for calling variants in pooled samples where the number of samples is not known. This is the exact scenario we have here: in our sample we have multiple mitochondrial (or bacterial or viral) genomes, but we do not know exactly how many. Thus we will use the
--pooled-continuous
option of FreeBayes to generate frequency-based variant calls as well as some other options