Confusion about LOW_QUAL variant

[robert-a-forsyth]

Hi,

We've been using ClairS-TO as part of our pipeline and wanted to clarify a confusing call on a particular variant. It was marked as LOW_QUAL in the output vcf file with a QUAL score of 0. However, looking at the reads I can't find an obvious reason for this. Could you provide some information on where this determination might be coming from?

below is an image of several samples which all have this variant and all of which are marked as LOW_QUAL. The variant reads have a mapping quality of 60 and both strands seem to be represented. Any insight would be great

[JasonCLEI]

Hi @robert-a-forsyth ,

In ClairS-TO, a variant call being marked as LOW_QUAL can result from several factors. It could be due to a low confidence score from the neural network’s somatic variant prediction, or it could be triggered by one of our post-processing filtering strategies.

Looking at your IGV image, while there appears to be a visible deletion, visual inspection alone is insufficient to determine exactly why the model or the pipeline decided to filter this specific call. To help us investigate this further, could you please send a mini BAM file along with the reference genome you used to lchen@cs.hku.hk? Please be assured that the data will be used solely for debugging and resolving this issue.

Lei