semEP is an edge partitioning approach that combines a data mining framework for link prediction, semantic knowledge (similarities) from ontologies, and an algorithmic approach to partition the edges of a heterogeneous graph.

For more information about semEP see:

Guillermo Palma, Maria-Esther Vidal and Louiqa Raschid. Drug-Target Interaction Prediction Using Semantic Similarity and Edge Partitioning. Proceedings of the 12th International Semantic Web Conference (ISWC 2013). Italy. 2014. (PDF)

• AUTHORS: list of contributors of the semEP project.
• doc: Documentation about semEP.
• LICENSE: GPL version 2.
• Makefile: builds semEP.
• README.md: this file.
• src: source code.
• test: datasets to test semEP
  • test/yamanishi: link prediction dataset proposed by Yamanishi et al. [1].
  • test/pawels: drug side-effect dataset proposed by Pauwels et al. [2].
• VERSION: software version.

[1] K. Bleakley and Y. Yamanishi. Supervised prediction of drug-target interactions using bipartite local models. Bioinformatics, 25(18):2397-2403, 2009.

[2] E. Pauwels, V. Stoven, and Y. Yamanishi. Predicting drug side-effect profiles: a chemical fragment-based approach. BMC Bioinformatics, 12:169, 2011.

GNU GENERAL PUBLIC LICENSE Version 2.

• GNU Colecction Compiler (GCC) or Clang.
• GNU make (make).

semEP has been tested on FreeBSD, GNU/Linux, and OS X.

Clean and generate necessary files:

$>make clean

$>make

semEP has 7 command line arguments and all them are mandatory. Here, mandatory means that without specifying this argument, the program won't work.

semEP command synopsis:

semEP <-l left threshold> <-r right threshold> <left matrix> <left vertices> <right matrix> <right vertices> <bipartite graph>

where:

• bipartite grap: file with the bipartite graph (BPG) to study.
• left threshold: threshold of similarity between the left side vertices of the BPG.
• right threshold: threshold of similarity between the right side vertices of the BPG.
• left matrix: similarity matrix file with the similarities between the left side vertices of the BPG.
• left vertices: file with the list of vertices of the left side of the BPG. The order of the vertices corresponds to the left matrix.
• right matrix: similarity matrix file with the similarities between the right side vertices of the BPG.
• right vertices: file with the list of vertices of the right side of the BPG. The order of the vertices corresponds to the right matrix.

• Compute the partitions in a drug-target bipartite graph and get predicted links on nuclear receptor dataset:

./semEP -l 0.3061 -r 0.1614 test/yamanishi/nr/nr_matrix_drugs.txt test/yamanishi/nr/nr_drugs.txt test/yamanishi/nr/nr_matrix_targets.txt test/yamanishi/nr/nr_targets.txt test/yamanishi/nr/nr_drug-target_graph.txt

• Compute the partitions in a drug-target bipartite graph and get predicted links on enzyme dataset:

./semEP -l 0.2174 -r 0.0198 test/yamanishi/e/e_matrix_drugs.txt test/yamanishi/e/e_drugs.txt test/yamanishi/e/e_matrix_targets.txt test/yamanishi/e/e_targets.txt test/yamanishi/e/e_drug-target_graph.txt

A bipartite graph is a graph whose vertices can be divided into two disjoint sets L and R such that every edge connects a vertex in L to one in R. We can denote a bipartite graph as G=(L,R,E), with E denoting the edges of the graph. We call the set L as the left side of the bipartite graph, and the set R as the right side of the bipartite graph.

The first line of the bipartite graph file, contains the number of edges. The following lines in the file shows the edges data, where each line corresponds to a edge and contains the node of the left side, the node of the right side, and the cost of the edge.

[number of edges n]
[right-node edge-1][TAB][lef-tnode edge-1][TAB][edge-cost-1]
..
..
..
[right-node edge-n][TAB][left-node edge-n][TAB][edge-cost-n]

Files with the matrices must contain the similarities between the vertices of the left side, or the right side. The file format is as follows:

[number of rows and columns]
[sim vertex-1 vertex-1][SPC]...[SPC][sim vertex-1 vertex-n]
..
..
..
[sim vertex-n vertex-1][SPC]...[SPC][sim vertex-n vertex-n]

The file format is as follows:

[number of vertices n]
[vertex 1]
..
..
..
[vertex n]

Where [vertex x] is the identifier of vertex in the left side or right side of the bipartite graph. The order of each vertex correspond to the position of the vertex in the similarity matrix.

semEP produces two outputs:

1. a directory with suffix -Clusters that contains the clusters with the edge partitioning of the bipartite graph. Each cluster correspond to a file on the directory;
2. a file with suffix -Predictions.txt that contains the semEP predictions for each cluster and the probability of each prediction.

I hope you find semEP an useful tool. Please, let me know any comment, problem, bug, or suggestion.

Guillermo Palma

palma at l3s dot de