Below are some R packages I've been exploring that implement methods described in the below review paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5784953/
R packages: graphite, ToPASeq, clipper, spia, topologyGSA, DEgraph, CePa
The ToPASeq package is totally overwritten by a new version that seems to be "in progress", as only the PRS function is available, and old GitHub repos for ToPASeq seem to be missing important code that would make the original package work as intended. I've contacted the authors with no response.
So, I moved on to the original implementations of the methods described in the review paper. Unfortunately, all of these methods are implemented around the graphite / KEGGgraph pathways, and none of them allow me to query a basic graph structure with a subset of nodes.
The main obstacle of these packages are that they require inputs from pathways from KEGG. I have a growing frustration with KEGG because of its very little coverage with my data. What if I'm not working with knowledgebases? What if I want to query a CUSTOM pathway/network, provided by myself?
I am seeking a GENERIC implementation of the methods described in the review paper, that takes a graphNEL object, igraph object or adjacency matrix and a subset of nodes as inputs, and outputs the respective scores / p-values, respective of the method used.
Does anyone have any experience with these methods and know more about how to use these methods outside the context of KEGG / curated pathways? I am even going so far as to try to write a custom KGML to "trick" these implementations into working with my custom networks I learned directly from data.