added background subtraction #48
Conversation
|
There was a problem hiding this comment.
Looks good to me! I would only change the exposure time to be a number to allow for floats as we have a dataset with such exposure times.
I'll be hopefully updating the nf-core module for Backsub soon so it adheres to nf-core standards more with conda support.
I'm leaning towards having a different step name than the tool name, so for the step name, I think it should be background/af_correction/channel_subtraction, with Backsub as the tool to do it. What are the preferences from your side?
| @@ -34,6 +34,15 @@ | |||
| "emission_wavelength": { | |||
| "type": "integer", | |||
| "errorMessage": "" | |||
| }, | |||
| "exposure": { | |||
| "type": "integer" | |||
There was a problem hiding this comment.
| "type": "integer" | |
| "type": "number" |
From what I see in other nf-core pipelines, modules are imported and used under their original names (unless a module is used multiple times and must be aliased for uniqueness) and publishDir folder names also generally follow the module name. Sometimes publishDir paths are grouped under a higher level folder if there are many alternatives for the same step like we have for segmentation. The mcmicro-legacy module names and publishDir layout did not use this style, rather naming everything by its purpose, e.g "registration", "segmentation", "quantification", and we did start to use that same publishDir structure here in mcmicro 2.0. I think we should pivot to adopting the nf-core style for consistency though. |
| @@ -19,6 +19,9 @@ params { | |||
| // Illumination correction | |||
| illumination = null | |||
|
|
|||
| // Background subtraction | |||
| backsub = false | |||
There was a problem hiding this comment.
Do we need this, or can it be inferred by the presence of the relevant columns in the marker sheet?
There was a problem hiding this comment.
Oh, yes, that's a good idea. We could trigger background subtraction on the presence of a remove column in the marker sheet.
Is there a better column to use?
There was a problem hiding this comment.
The column critical for the module would be background, as only using the tool for channel removal seems like an unnecessary data duplication step.
There was a problem hiding this comment.
ok, sounds good, I'll use the background column. Thanks!
I thought it might be used for something else as well, but it seemed like the remove column would only be used for background subtraction.
There was a problem hiding this comment.
Just to add here, it might be good to check for both the presence of the background and exposure columns as both are required for channel subtraction. If you check by background, and no exposure column is given, the tool would give an error.
I think the best way to approach this would be to: check for presence of the background column, if it is given, check the values in the column, if none are a string, we skip Backsub. Otherwise, check for presence of the exposure column (and values), if not provided, an error should be thrown, and otherwise Backsub is run.
There was a problem hiding this comment.
Should validate that background column values are a subset of the marker_name values in validateInputMarkersheet.
| input[0] = Channel.of( | ||
| [ | ||
| [id:"TEST1", cycle_number:1, channel_count:4], | ||
| "https://raw.githubusercontent.com/nf-core/test-datasets/modules/data/imaging/background_subtraction/cycif_tonsil_registered.ome.tif", |
There was a problem hiding this comment.
I don't think this is an actual background-subtraction test case, but structurally I guess it's fine.
| channel_number + "," + cycle_number + "," + marker_name + "," + exposure + "," + background + "," + remove}] } | ||
| .flatten() | ||
| .map { it.replace('[]', '') } | ||
| .collectFile(name: 'markers_backsub.csv', sort: false, newLine: true, storeDir: "${workDir}") |
There was a problem hiding this comment.
Don't use storeDir here. collectFile returns a channel, and you should use that channel directly where needed downstream. Check out ch_mcquant_markers and its usage in calling MCQUANT above.
| .flatten() | ||
| .map { it.replace('[]', '') } | ||
| .collectFile(name: 'markers_backsub.csv', sort: false, newLine: true, storeDir: "${workDir}") | ||
| BACKSUB(ASHLAR.out.tif, [[id:"$ASHLAR.out.tif[0]['id']"], "${workDir}/markers_backsub.csv"]) |
There was a problem hiding this comment.
As mentioned above, instead of an explicit path for the csv, use ch_backsub_markers. You'll need to use combine to join the collectFile-returned single-value channel with each entry in the meta-map channel for the second arg to BACKSUB.
I was under the impression our goal is to still keep the multiple choice aspect based on steps - especially when it will come to defining the "execution stream" - e.g. staging, registration, AF_correction, segmentation and quantification. The broader steps are defined and used to determine the "stream", but for each step (or several steps), we would allow for multiple options (e.g. staging, registration*, segmentation). I think it would be somewhat confusing to start mixing process names into the broader execution stream which is why I made the point above. |
PR checklist
nf-core lint).nextflow run . -profile test,docker --outdir <OUTDIR>).nextflow run . -profile debug,test,docker --outdir <OUTDIR>).docs/usage.mdis updated.docs/output.mdis updated.CHANGELOG.mdis updated.README.mdis updated (including new tool citations and authors/contributors).