MoProEmbeddings

Python implementation of the moment propagation (MoPro) embeddings described in Prediction of cancer driver genes through network-based moment propagation of mutation scores (Anja C. Gumpinger, Kasper Lage, Heiko Horn and Karsten Borgwardt). See https://academic.oup.com/bioinformatics/article/doi/10.1093/bioinformatics/btaa581/5861532 for the publication.

Requirements

To execute the python package, the following modules are required.

numpy
scipy
python-igraph
pandas

Tests

Unittest can be found in ./tests, and executed with ./tests/unittest/runAllTests.sh.

Moment Propagation Embeddings.

The generation of moment propagation embeddings is a four-step process:

1. k-hop path weights have to be computed (functionality in path_weights.py)
2. The data have to be represented in an igraph object (functionality in basegraph.py)
3. Neighborhood features are created using the igraph representation (functionality in features.py)
4. MomProp embeddings are generated (functionality in data.py)

Example.

IPython notebooks that show how to execute the above steps can be found in ./examples. First, the data has to be processed as shown in ./examples/preprocessing.ipynb. Next, the MomProp embeddings can be computed, as explained in mom_prop_embeddings.ipynb. In case the notebooks are not propely displayed, we recommend https://nbviewer.jupyter.org/.