Add X and Y chromosomes to CNV segment to gene symbol mapping

[jaclyn-taroni]

Purpose/implementation Section

Here, I'm adding handling of X and Y chromosomes to the focal-cn-file-preparation module.

• I added steps to the CNVkit file prep that made the status column use similar logic to what is in the ControlFreeC file.
• I added a step for merging overlaps specifically for the X & Y chromosomes into 01-prepare-cn-file.R. The autosome and sex chromosome files get saved separately. The sex chromosome files include the germline_sex_estimate column from pbta-histologies.tsv.
• I've added steps to compress the X&Y TSV files.

What GitHub issue does your pull request address?

#186

Directions for reviewers. Tell potential reviewers what kind of feedback you are soliciting.

Which areas should receive a particularly close look?

The sex chromosome loss, gain, neutral calls in the CNVkit preparation steps (00-add-ploidy-cnvkit) could use a good look. I am trying to mirror what logic is in the ControlFreeC file.

Is there anything that you want to discuss further?

At this point, I've chosen to not mark amplification in X and Y chromosomes and I've left in the neutral calls because I feel that a downstream analyst might want to give these a careful look.

Is the analysis in a mature enough form that the resulting figure(s) and/or table(s) are ready for review?

N/A

Reproducibility Checklist

This has already been added to Docker and CI at an earlier point.

[jashapiro]

This looks good, the only real question I have is about the logic of calling gain and loss in males. I also was curious about the logic of skipping XY processing for controlfreec only in the run script.

[jashapiro]

No xy option here? So the cnvkit files always do xy? Is there a particular reason for that?

[jaclyn-taroni]

The default is to run the X and Y steps, so I do not include it here. The reason an option is included for ControlFreeC is to allow this to pass CI. My suspicion is that the ControlFreeC subset files don't represent the X and Y chromosome enough, causing an error with IRanges::mergeByOverlaps and I still want the X and Y portion of the code to be tested with CNVkit.

[jashapiro]

I wonder if these are the right cutoffs... I'm thinking that if tumor_ploidy is 3, then the expected number of X or Y chromosomes is 1 or 2, not 1.5; so everything would be categorized as gain or loss. If we are within 0.5 copies of the 0.5 * tumor_ploidy, perhaps we should have a different call?

[jaclyn-taroni]

Looking at ControlFreeC calls for X and Y chromosomes only in instances where germline_sex_estimate == "Male" and grouping by chromsome, copy number, status, and ploidy:

chr	copy.number	status	tumor_ploidy	n
X	0	loss	2	176
X	2	gain	2	1447
X	3	gain	2	255
X	4	gain	2	148
X	5	gain	2	117
X	6	gain	2	74
X	7	gain	2	27
X	8	gain	2	17
X	9	gain	2	13
X	10	gain	2	5
X	11	gain	2	2
X	12	gain	2	3
X	13	gain	2	7
X	14	gain	2	2
X	16	gain	2	1
X	17	gain	2	1
X	19	gain	2	1
X	21	gain	2	1
X	24	gain	2	3
X	27	gain	2	1
X	28	gain	2	1
X	31	gain	2	2
X	34	gain	2	1
X	36	gain	2	1
X	53	gain	2	1
Y	0	loss	2	299
Y	2	gain	2	3338
Y	3	gain	2	1026
Y	4	gain	2	201
Y	5	gain	2	56
Y	6	gain	2	14
Y	7	gain	2	10
Y	8	gain	2	3
Y	10	gain	2	5
Y	11	gain	2	1
Y	17	gain	2	1
Y	19	gain	2	1
Y	30	gain	2	1
Y	41	gain	2	1
Y	54	gain	2	1
X	0	loss	3	17
X	1	loss	3	1445
X	2	gain	3	3008
X	3	gain	3	166
X	4	gain	3	45
X	5	gain	3	50
X	6	gain	3	35
X	7	gain	3	20
X	8	gain	3	17
X	9	gain	3	18
X	10	gain	3	8
X	11	gain	3	4
X	12	gain	3	1
X	14	gain	3	2
X	15	gain	3	1
X	46	gain	3	1
Y	0	loss	3	42
Y	1	loss	3	1094
Y	2	gain	3	1538
Y	3	gain	3	435
Y	4	gain	3	212
Y	5	gain	3	56
Y	6	gain	3	22
Y	7	gain	3	7
Y	8	gain	3	5
Y	9	gain	3	3
Y	10	gain	3	5
Y	12	gain	3	1
X	0	loss	4	9
X	1	loss	4	311
X	2	neutral	4	1412
X	3	gain	4	583
X	4	gain	4	65
X	5	gain	4	30
X	6	gain	4	21
X	7	gain	4	18
X	8	gain	4	11
X	9	gain	4	12
X	10	gain	4	6
X	11	gain	4	7
X	12	gain	4	7
X	13	gain	4	6
X	14	gain	4	6
X	15	gain	4	2
X	17	gain	4	3
X	18	gain	4	2
X	21	gain	4	1
X	24	gain	4	1
X	25	gain	4	1
X	27	gain	4	1
X	28	gain	4	1
X	39	gain	4	1
X	53	gain	4	1
X	108	gain	4	1
Y	0	loss	4	61
Y	1	loss	4	236
Y	2	neutral	4	534
Y	3	gain	4	366
Y	4	gain	4	181
Y	5	gain	4	134
Y	6	gain	4	87
Y	7	gain	4	24
Y	8	gain	4	23
Y	9	gain	4	10
Y	10	gain	4	4
Y	11	gain	4	3
Y	12	gain	4	2
Y	18	gain	4	1
Y	19	gain	4	1

I believe the logic I've implemented for CNVkit matches this.

[jashapiro]

It does appear so. Something to keep in mind downstream though. 🤨

[jaclyn-taroni]

Really missing that 💯 reaction option here

[jashapiro]

With the caveat that I don't like how ControlFreeC handles sex chromosomes in males when ploidy is odd, this looks good to go.