ScPred - version inconsistency for feature selection

[hexhowells]

I am building a workflow in Galaxy using scPred and have encountered an issue with the latest version (1.0.2+galaxy0) of the tool ScPred feature space, when running the tool i get the following error:

Error in getFeatureSpace(data_seurat, pvar = opt$prediction_column, correction = opt$correction_method,  : 
  Invalid class for object: must be 'Seurat'
Execution halted

It seems the tool only accepts Seurat objects, however all prior tools that should be run output a SingleCellExperiment or ScPred object.

When using a previous version of the tool (1.0.1+galaxy0) it runs fine with no errors, as I believe prior versions accept Seurat and ScPred objects.

---

Also to note the prior tools that get run (Scpred pre-process/Scpred train-test split and Scpred eigen-decompose) all have the latest version as 1.0.1+galaxy0. So i'm guessing either the updated version has broken something or only some of the tools have been updated?

A Galaxy workflow reproducing the error can be found here. Step 13 uses the latest version of the ScPred feature space tool whereas 29 uses the latest version.

[pcm32]

@hexhowells I have managed to merge changes on scPred-cli to make the continous integration test work. I have added Seurat and loomexperiment as depenedncies for the testing process, so the SingleCellExperiment objects with Seurat or loom flavour should work once we make a new release of scpred-cli. Towards this end, do you have any other needs that we need to incorporate on scPred-cli before releasing it? If that is so, could you PR here please? Thanks!