-
-
Notifications
You must be signed in to change notification settings - Fork 62
[WIP] CNVkit tool definitions #93
base: master
Are you sure you want to change the base?
Conversation
tools/cnvkit-batch.cwl
Outdated
| log-CNR of chrX; otherwise male samples would have -1 chrX). | ||
| inputBinding: | ||
| position: 2 | ||
| prefix: --male-reference |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
CWL tip: for Argparse only the position dependent arguments need their position specified. Arguments that have a prefix like --male-reference can occur in any order, so it would be nice if cwlargparse didn't specify the unneeded positions in these cases.
|
I think it is fine to just check in the generated descriptions; don't worry about writing a specific test. As long as the generated output parses, that is good enough for now. |
|
I'm totally with @mr-c, we should focus first on CWL, not on specific tools since the amount of work can be quite substantial. If you want to see whether one of the CNVkit subtools works it's fine to dedicate some focused effort, but by no means aiming to cover the whole suite of tools. Hope that makes sense ;) |
|
OTOH, for a good example on how to test different tools (in my case SV callers), MetaSV has it quite well wrapped up: https://github.com/bioinform/metasv But this is just an example, don't spend too much time looking through it. |
|
@brainstorm , that's great! I misinterpreted the goal of the PR, it was not to pass Travis checks but to merely validate those tools. In that case, I'll fix the job file (@mr-c pointed indirectly on that issue) and push all other tools. UPD. I should have looked more closely at |
tools/cnvkit-docker.cwl
Outdated
| ################################################################# | ||
|
|
||
| FROM python:2.7 | ||
| MAINTAINER Anton Khodak <[email protected]> |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is great! Thanks for wrapping this on a docker container 👍
|
Hi guys, I'm happy to help with testing CNVkit and/or tweaking the test suite to play better with argparse2cwl. You can skip wrapping anything marked "deprecated" (e.g. |
|
@etal, very happy to have you help Anton with that. I was looking at the |
| prefix: --diagram | ||
|
|
||
| outputs: | ||
| [] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There are several outputs from this command and they vary based on the input BAM filenames and the options given.
- For each tumor/test-sample BAM named e.g. Sample.bam, the outputs are: "Sample.targetcoverage.cnn", "Sample.antitargetcoverage.cnn", "Sample.cnr", "Sample.cns"
- If the
--scatteroption is given, then for each tumor/test sample, "Sample-scatter.pdf" is created - Similarly, the
--diagramoption creates "Sample-diagram.pdf" - For all of the above, if
-d/--output-diris specified, the created file names are relative to (i.e. in) that specified directory - If the
-r/--referenceoption is not given, then a.cnnfile is created either with the filename given by--output-reference(regardless of the-d/--output-dirpath) or by default "cnv_reference.cnn"
|
I've released a new minor version of CNVkit that drops the deprecated parts and introduces a few new options. I think the current CWL wrappers in Anton's repo should still work, but |
Standing PR to add tool descriptions (created by argparse2cwl) and tests for CNVkit tools .
Issues I encountered on first steps:
I think this error might be caused by irrelevant data.
Also, I couldn't find copy number reference profile sample files (.cnn) at all. If somebody who uses CNVkit frequently could give me a hint where to take proper data, my work in testing would have been much facilitated.
tool-name-test.yamlfile format is specified. It was intuitively understandable what to write there, but I wish somebody pointed the standard for those files.