Thanks @blessyantony9! I see you copied the PageRank.py to src/FastSinkSource/src/algorithms, and also added the PathLinker repo to src. It looks like the PathLinker repo isn't used anywhere, is that right? Can we remove it?

Thank you @jlaw9 for reviewing. Yes, the aim is to reuse PageRank from within PathLinker which is added as a sub-repo in this repo. However, unfortunately, there is version incompatibility for one of the packages between PathLinker and SARS-CoV-2-network-analysis: networkX. In PathLinker it is v1.11 whereas in the current repo it is v2+. So, until we upgrade the version of this package in the whole of PathLinker, I have copied the code of PageRank over to src/FastSinkSource/src/algorithms and updated the usages only in this file as a tactical solution. I have included this as a TODO comment in the file.
I hope to update the PanthLinker repo soon and get rid of this redundant file. Hence, I have retained the sub-repo but can definitely remove it and wait until the update to networkX v2+ is complete before pulling it in.

Ah ok I see. I appreciate the work you did to incorporate the networkx graphs to be able to use that PageRank implementation.

Since we already have the other algorithms implemented in terms of matrix-vector multiplication, I think it would keep the code simpler (and faster than networkx) to do the same for RWR i.e., s^(i+1) = aPs^(i) + (1-a)y. And then we don't have to worry about the networkx versions.

I had forgotten that I had added such an implementation to the annotation_prediction repo, which is the updated version of the FastSinkSource repo (I had planned to update the FastSinkSource part of this repo to use that annotation_prediction repo, but hadn't gotten around to it unfortunately).

It looks like there are a couple of features that PageRank.py code has that my rwr.py doesn't have, such as the ability to use non-uniform weights for the restart probabilities, but I think that would be easy enough to add if we we wanted to use that.

Thank you @jlaw9 for the inputs about using matrix multiplication. I was unaware of the existing matrix implementation. I too implemented RWR using sparse matrices for another project within the N3C Enclave (Palantir) which too assumes default weight=1, but as you mentioned can be tweaked to incorporate non-uniform weights. I'll discuss the available matrix implementations with Dr.Murali and update this pull request accordingly (if we decide to go with the matrix version).

@blessyantony9 please go ahead and use the matrix implementation from the other project here as well.

Thank you @tmmurali. I'll update the code to use RWR from the annotation_prediction repo.

@jlaw9 As discussed, I've added the PageRank Matrix implementation. Apologies for the delay. Please let me know if you have any questions or if any changes are required.