Modules across all cell types

[sandcell10]

Hi @smorabit

Thank you for your help and the awesome package! I know part of the point of hdWGCNA is to perform co-expression network analysis in individual cell types. However, is it valid to get an idea of co-expression in an entire data set by performing the analysis across all cell types (specifying all cells in the SetDatExpr command)? Are there any advantages/disadvantages to grouping cell types this way?

Also, I noticed that the number of modules created can differ pretty substantially between different cell types. Additionally, if I use all the cell types as described above, sometimes only a few modules are created. Would you be able to comment on why this is?

Any help would be appreciated and thank you again!

[smorabit]

Hi @sandcell10,

is it valid to get an idea of co-expression in an entire data set by performing the analysis across all cell types (specifying all cells in the SetDatExpr command)?

hdWGCNA is flexible in terms of what cells you provide as input. For example, one cell type, the whole dataset, one cell cluster, one disease group, etc. Any of these groupings are "valid", but your downstream interpretations will change depending on your inputs.

Also, I noticed that the number of modules created can differ pretty substantially between different cell types. Additionally, if I use all the cell types as described above, sometimes only a few modules are created. Would you be able to comment on why this is?

Unfortunately I don't have a definitive answer or any meaningful analysis investigating why some cell types have more or fewer modules detected.