GENIE

artifact_finder.py

The basic principle is first to simply filter unique variants for which

• counts in a given panel >= 10

• counts aggregated from all other panels is < 10 and there are >= 3 different panels that cover the variant in question

Of course, we can adjust these thresholds. In particular we need better account for truly different panels from different sites vs different versions of the same "panel" from the same site. And, we will more than likely need some filtering on number of samples per panel.

I also annotated the resulting set with a fisher p-value for the variants that pass this filtering, with the 2 x 2 table of

Number of variants called in panel	Number of samples for this panel
Number of variants called in other panels with coverage for this variant	Number of samples for panels with coverage for this variant

genie_fusion_check.py

• catalogues the number of samples with fusions calls in the data_fusion.txt file of a given release and aligns this against what the data_clinical_sample.txt table says in conjunction with the assay_information.txt.