Open ihodes opened 9 years ago
I'm not sure I agree with this. One use case for CycleDash is to tune thresholds and quickly see what effect they have on precision/recall. If you added sample::RD > 100
as a filter, I'd expect to see the precision increase and the recall go down. Same for quality > 0.99
or some such. The numbers we'd report with those filters are still meaningful.
I don't think we're disagreeing; I'll retitle the issue & add to the top comment to make it clearer.
Filters other than reference/alternates/sample_name are ignored for the comparison VCF—it would be nice if we have the ability to filter JUST the variants in the selected run or just the variants in the (any of the) comparison runs (https://github.com/hammerlab/cycledash/issues/692) and if we could be more explicit about what's being joined on in the comparison (https://github.com/hammerlab/cycledash/issues/693) so that the statistics are more meaningful and clear (https://github.com/hammerlab/cycledash/issues/691).