Open PengZhangJHU opened 1 year ago
Hi Peng,
More details about the false-positive identification strategy can be found in the paper. As you can see, BF and other tool-related output are not considered. Of course, you can use BF or other values to further refine your filtering.
I found that in my results, the numbers of h.hm.variants+n.ht.discard.CNV+h.ht.confirm.CNV = n.total.variants for most duplication calls or close. But for deletions, I have cases like n.total.varaints=67, whereas n.hm.variants=55 and no numbers in n.ht.discard.CNV+h.ht.confirm.CNV, so they don't add up. Why is the difference? Is n.total.variants (67) here all variants in the deletion region, but only 55 of 67 pass the threshold [min.total.depth threshold (20) and also not within margin.pct of 10], or something else?
In addition, I see in the example CNV dataset (DECoN.CNVcalls.csv), there are BF, reads.ratio information. We used ExomeDepth for our whole exome sequencing which also have those information. Do you use them to filter the CNV call besides CNVfilteR results? Any suggestions on how to apply them together to filter CNV calls? Thank you! -Peng