SRW-PPINetworks

This repository is an implementation of the Supervised Random Walks algorithm (MATLAB original code link) for Protein-Protein Interaction Networks. The algorithm is implemented in Python where the optimization is implemented using the function fmin_l_bfgs_b from Scipy module with Wilcoxon-Mann-Whitney (WMW) loss function. Learning the parameter vector w can be done with the function supervised_random_walks, and with the parameter vector w, the function random_walks gives the random walk parameter vector p. The alternative implementation of the algorithm is implemented to work on GPU using the NDArray API of MXNet.

Scipts:

• supervised_random_walks.py - implementation of the SRW on CPU
• supervised_random_walks_gpu.py - implementation of the SRW on GPU
• train.py - train and test the SRW algorithm
• Preprocessing scripts:
  • find_largest_component.py
  • min_max_normalization.py
  • semanticSimilarityGoTerms.R and semanticSimilarityGoTerms.py - calculation of semantic similarity of GO terms
  • semanticSimilarityProteins.R and semanticSimilarityProteins.py - calculation of semantic similarity of proteins