scatter finds zero probes when using regions

[ViktorHy]

I have a problem with version 0.9.6. I can plot single chromosomes using the following command:

cnvkit.py scatter -s Sample.cn{s,r} -c 17 -o 17.png <- this works

But if I want to plot I defined area say

cnvkit.py scatter -s Sample.cn{s,r} -c 17:37000000-38000000 -o 17.png <- this doesn't

Showing 0 probes and 1 selected genes in region 17:35999999-39000000

Similarily if I try to plot only a selected gene.

cnvkit.py scatter -s Sample.cn{s,r} -g ERBB2

Showing 0 probes and 1 selected genes....

I tried with an older installation, version 0.9.1 and all the above commands works just fine

[tetedange13]

Hi,


Had the same bug (CNVkit v0.9.6) and it seemed to come from a problem in _irange_simple() at this line:

cnvkit/skgenome/intersect.py

Line 165 in f4ccc90

yield (slice(start_idx, end_idx), start_val, end_val)

Possible fix:
yield (slice(table.index[start_idx], table.index[end_idx]), start_val, end_val)

Explanation (guessed):
The start_idx and and end_idx in the slice yielded are actually indexes relative to a subset of "table", created within idx_ranges():

cnvkit/skgenome/intersect.py

Line 121 in f4ccc90

table = table[table['chromosome'] == chrom]

The problem is that coming back to iter_ranges(), this slice is applied to the whole "table" (full size not subset), to yield a "subtable" object. The indexes being still relative, we catch a wrong region (often in chrom 1).

IMPORTANT EDIT:
My quick fix seems to bug during segmentation, as the "table" given is already subsetted by chromosome (so internal indexes have to be yield, no need to convert them back to "not internal" one).

Hope it helps. Best regards. Felix.