Pathway level analysis of single-cell gene expression

I'm looking for single-cell specific methods to construct (using gene expression data) new features that express pathway "level" or "activity", and then use these for clustering cells.

One example for bulk RNA-seq is PLAGE, implemented in the GSVA R package.

I have found only one method for single-cell RNA-seq that comes close: pagoda in the scde R package. This performs PCAs for each pathway, using its genes.

A very similar question is Classification (supervised learning) of expression data on pathway level, on bulk/microarray data.

• Hi, I find a bit unclear your question, could you clarify if you want to find (new?) pathways in scRNA-seq or how to analyze scRNA-seq using pathways? If the later, what is the problem with the methods you mention?
– llrs
May 24 '18 at 7:09
• I'm not trying to find new pathways, I want to use pathway- or biological function-level information to classify or understand cell diversity. GSVA is designed for bulk data, so I don't yet know how applicable it is to single cells. scde might be good. May 25 '18 at 10:43
• 1)Pathways might (most probably do) change for each cell line (and in fact if it is not from a isolation, you don't know if this is a different cell line or not unless you traced it somehow) 2) We don't know enough about pathways and how they work (ie, there is few reliable information available, different databases define different what a pathway is, either by interactions, with kinetic constants an thus by contact (but then in which conditions are these measured?))
– llrs
May 25 '18 at 10:54
• Thanks for pointing me to scde. I found that it identifies "known pathways or novel gene sets"
– llrs
May 25 '18 at 10:59