initial commit of dreamlet function

[erflynn]

Initial attempt to set up dreamlet wrapper. Takes as input a pseudobulked count matrix and metadata.
I tried to follow conventions from Dan's examples where possible.
To do items:

• add info about the default processAssays() behavior (which also filters based on min.cells)
• add back in the dreamlet run df
• update function to take ... and/or additional specific params to pass to dreamlet
• add additional documentation (@dtm2451 I'm not sure what you generate the .txt file with?)
• add example usage
• additional testing

[dtm2451]

I make the txt file with document::document(), and the call was: document::document('single_cell/pseudobulk_function.R', output_directory = 'single_cell/', check_package = FALSE).

[erflynn]

Added some updates to the input check functions as well as updated dreamlet (apologies, this should probably be two separate PRs but I wanted to do both)

Some notes to discuss:

• nomenclature:

  • contrast --> contrasts
  • do we want cell_type or celltype?
  • add gene name to the output table (I dont like having as rownames)

• when to do input checks?

  • for contrasts, this is easier at the model stage, so I did it here but maybe we should do earlier?
  • for within cells -- a lot of this waits until the .input_check function is first called in the cells group, is this ok?
  • I added .input_check_dge_case_ref() since it's needed for within cells and also the general one

• what should we do about NA values in dge_by (or other) columns?

• what to do with numeric values as dge_by? currently we get a warning they will be set to factor, but we may not want to do this (unless case_group and reference_group are 0 and 1 or something)

• do we want to use MIN.COUNT (minimum number of counts required per sample when calculating whether a gene should be retained (default = 5)) more generally? Needed for dreamlet

• where to put additional small helper functions?

[dtm2451]

• nomenclature:

  • contrast --> contrasts

I think I like the plural form better! Because the functionality it offers is for grabbing results from multiple directions -- like if you are only looking at one comparison (one contrast) you might as well just use case_group and ref_group inputs.

• do we want cell_type or celltype?

Seems like we decided on celltype previously, here

• add gene name to the output table (I dont like having as rownames)

I definitely agree about this. @ravipatel4? We should add it to the same list this list of column names if so!

[dtm2451]

• when to do input checks?

  • for contrasts, this is easier at the model stage, so I did it here but maybe we should do earlier?
  • for within cells -- a lot of this waits until the .input_check function is first called in the cells group, is this ok?
  • I added .input_check_dge_case_ref() since it's needed for within cells and also the general one

• what should we do about NA values in dge_by (or other) columns?

• what to do with numeric values as dge_by? currently we get a warning they will be set to factor, but we may not want to do this (unless case_group and reference_group are 0 and 1 or something)

• do we want to use MIN.COUNT (minimum number of counts required per sample when calculating whether a gene should be retained (default = 5)) more generally? Needed for dreamlet

The rest of these questions I don't feel as comfy just saying my piece and us treating that as the call, so maybe we should, as you suggested live already, try and have another "dge-working-group" meeting soon to decide!