Alternative graph data structures

[zietzm]

We are considering creating another base representation of hetnets. One of the main goals is to facilitate faster network loading, which at present can take over a minute and a half to load a graph.

The following are under consideration:

• Mambo, which relies on SNAP.
• GraphFrames, a graph package for Apache Spark. Advantages of this framework are that metadata and multiple edge types are supported in a logical way, Spark scales well in case we eventually make hetmech a back-end server application, and the data structure is highly-used. However, we would have to re-implement the DWPC in a way that either computes exact path-level results (path and corresponding DWPC) more slowly than the current matrix implementation or computes path-level DWPC only.
• HetMat, what would be an internally-produced matrix-first representation of the network. This option would allow us to store the entire Hetionet-v1.0 and five permutations in about 30 MB as scipy sparse matrices in .npz format. Moreover, unlike the other methods, we would likely not have to change more than a few functions in order to load on-disk adjacency matrices.

[dhimmel]

Another package of interest could be xarray, which provides "N-dimensional variants of the core pandas data structures." I think it could be ideal for matrix based representations of hetnets. However currently it can only use numpy arrays as its backend and lacks scipy.sparse support. Therefore it's probably not appropriate at this time.

Check out 2.xarray.ipynb where we encode Hetionet v1.0 as an xarray.

[dhimmel]

The following products are related and may be good for storing dataframes and potentially matrices on disk:

• https://github.com/wesm/feather
• https://github.com/apache/arrow
• https://parquet.apache.org/

[dhimmel]

For storing nodes for hetmech, I'm thinking we should use sqlite, to enable fast lookup of node positions from names. However, not sure if we should do this now in #97 or later.

[zietzm]

Closed by #97

[dhimmel]

multinetx

I recently came across multinetx (GitHub), which is a:

python package for the manipulation and visualization of multilayer networks. The core of this package is a MultilayerGraph, a class that inherits all properties from networkx.Graph().

I don't think we have any use for this package at the moment, but I wanted to note it here, so we can keep an eye on its development.