Unused parameters in Scmap tools

[hexhowells]

Some scmap tools in Galaxy have parameters which don't seem to be utilised. Additionally there are some parameters which could be included which can help get the best results from the tools.

Select features

Scmap select features includes the parameter Number of features which looking in scmap-cli (here) isn't being used in any of the code.

From the scmap documentation the parameter n_features should be used in the following way:

selectFeatures(
    object, 
    n_features=500, 
    suppress_plot=FALSE
)

Cluster projection

Scmap cluster projection includes the parameter threshold which also seems to not be utilised in the scmap-cli tool here

Looking at the scmap documentation for that function the threshold parameter is used in the following way:

scmapCluster(
    projection, 
    index_list, 
    threshold=0.7
)

Cell projection

The Scmap cell projection tool could also include two additional parameters which aren't required but would aid in refining the results of the annotation. Specifically the function scmapCell2Cluster function (found here) in the scmap-cli tool can take two additional parameters:

• w an integer specifying the number of nearest neighbours to find
• threshold the threshold which the maximum similarity between the query and a reference cell must exceed for the cell-type to be assigned

These parameters can be found in the documentation here

---

For context I am running these tools in a Galaxy workflow on the usegalaxy.eu server.

[pcm32]

@hexhowells unfortunately all of those tools have no one with funding to support them :-( and I left the EBI nearly a year ago (where we had some involvement with them). I keep supporting the scanpy part there, but not the cell mapping part. What I could offer is that if you attempt a PR on that scmap-cli with your changes and the test pass without infinite re-arrangements of its conda dependencies, we could make a new release and then, if you help us with the few galaxy wrapper changes and they keep passing tests on the new scmap-cli release, we could push them to the toolshed via our CI.

I did try to make changes on scPred-cli to be able to help you on other issues (on my own time - I have no funding for this) but as you can see certain things with conda are not playing ball :-( there (see ebi-gene-expression-group/scpred-cli#22).

[hexhowells]

@pcm32 thank you for looking into it, I have been working on making a fix and have just made two pull requests fixing the issues I have raised here:
(ebi-gene-expression-group/scmap-cli#14)
(ebi-gene-expression-group/scmap-cli#15)

The fixes are quite simple so hopefully they wont require much to get merged and working but I am not familiar with the process of wrapping them so I'm not sure if that will present issues.

---

Also as a note; when running the scripts locally to test I needed to change the lines here and here as they were causing errors, Since the tools are working in Galaxy currently I wasn't sure if it was something to do with my local setup so I haven't modified them.

[pcm32]

Thanks for those @hexhowells . I think that doing this is going to hurt, because I see that those repos are using still travis, which no longer work. We will need to transition those to github actions.

[bgruening]

@pcm32 @hexhowells can this one be closed?

[hexhowells]

Yes! This has now been resolved here: ebi-gene-expression-group/container-galaxy-sc-tertiary#318