CNV annotation

[maximus3219]

I have two comments/questions concerning CNV annotation by PCGR.

First:
For annotation of gene cross two adjacent segments, I think the current algorithm of PCGR annotation is to simply annotate that gene according to the result of the first segment. For example if first 20% of gene X overlaps with hetdel segments, and the latter 80% of gene X overlaps with dup segments, the gene X is labelled as hetdel.
I think it is better to assign the CNV state of these genes based on the majority overlap or weighted average of segment properties covering the gene.
We can also flags this gene as "complex" or "partial" events or flag intragenic breakpoints.

Second:
Currently the VARIANT_CLASS is annotated as follows:
gain: total copy number >= user-defined threshold; homdel - total copy number equal to zero; hetdel - total copy number equal to one; undefined other copy number states
With the above criteria, major CN = 1 and minor CN = 1 is labelled as undefined, which should be labelled as "neutral".
Please consider finer annotation like LOH and DUP-LOH, since you have info of major and minor CN.

[sigven]

Dear @maximus3219 ,

Thanks so much for this input! I believe both of them are very valuable suggestions on how the CNA annotation could be further improved. For your information, we are also having discussions as to how we can provide additional support to explore amplifications/losses relative to the ploidy (rather than hard thresholds, as is the only parameters configurable now). And I agree regarding the LOH part, this is currently very much underutilized in the annotation/report.

I've added your points to the development list. Thanks again!

best,
Sigve