Variant detection occurs in multiple stages.
1) All the potential variants in the original read alignment are identified, and are subsequently evaluated to remove likely FP while retaining high-confidence variants.
2) A closer investigation considers the reads supporting the variants, including information about other variants in the genomic neighbourhood (e.g. by re-aligning the reads with larger gaps than allowed in the original alignment).
– In some cases, an INDEL can result in FP SNVs (e.g. if the reads do not span the entire INDEL). The SOPHiA Platform identifies these FP artefacts and shifts the supporting reads accordingly (thus moving them from the FP to the real variant). In this situation, a variant can end up having 0% VF as the reads supporting it in the original alignment were shifted to another variant.
– FP artefact variants are kept in the SOPHiA Platform to show you that they were found in the original BAM alignment file, though further analysis showed there is little support for them. Otherwise, you may think they were missed by the SOPHiA Platform, especially if they were detected with a different software. So, in these cases the 0% variant fraction can be interpreted as an FP artefact.