Variants flagged as PONArtefact despite absence from gnomAD and SAGE PON

[silkrp]

Hi,

We are analysing matched WES data from a longitudinal HGSOC cohort with the following sample types per patient: Pre-lesion STIC tissue, primary HGSOC tumour, metastatic HGSOC tumour, and a matched normal.

Given the tumour type, we would expect to call TP53 variants across all tumour samples. However, in the majority of the patients, TP53 variants are being filtered out as PONArtefacts.

For example, the variant chr17:7675236 A>T is being annotated with PONArtefact in the SAGE filter column, so would be dropped if filtering only for PASS variants. However, after checking the gnomad and SAGE PONs, this variant doesn't seem to appear in either?

I'm wondering why this variant is being annotated as a PONArtefact despite not being in the PONs?

All the best,
Ryan

[silkrp]

Hi,

We've now realised that the PONArtefact annotation is a result of the PON generated during the training steps for targeted sequencing. It seems odd that when running the training only tumour samples are accepted, as I'm guessing that any variants that are highly prevalent across the training samples - such as TP53 in HGSOC - will be included in the PON, which then subsequently will get annotated as PONArtefacts by SAGE?

All the best,
Ryan

[scwatts]

Hi @silkrp, this sounds right to me. However, I've only developed oncoanalyser rather that the underlying tools themselves, can I suggest putting your question to the tool authors over at hartwigmedical/hmftools?