This software implements the PathLinker algorithm for reconstructing signaling pathways from protein interaction networks.
- Christopher L. Poirel ([email protected])
- Anna Ritz ([email protected])
- Nicholas Sharp ([email protected])
- T. M. Murali ([email protected]) [corresponding author]
PathLinker is a method for reconstructing signaling pathways from protein interaction and regulatory networks. Consider a protein-protein interaction network represented as a directed graph. The network may contain physical, signaling, and regulatory interactions between pairs of proteins. Given a query composed of a set of receptors and a set of transcription factors in this network that represent the "sources" and "targets" of some specific signaling pathway (e.g., the Wnt pathway), PathLinker attemps to compute a high quality reconstruction of this pathway. Specifically, PathLinker computes a subnetwork that connects the receptors in the query to the transcription factors in the query. In its output, PathLinker ranks the interactions in the network with respect to the query, as described below.
The primary algorithmic component of PathLinker is computing the k-shortest simple paths in the network from any receptor to any transcription factor. PathLinker accomplishes this task through a novel integration of Yen's algorithm with the A* heuristic, which allows very efficient computation for very large k values, e.g., 10,000, on networks with hundreds of thousands of edges. PathLinker ranks each interaction in the network by the index of the first path in which it appears.
PathLinker supports several different options for weighting the network. If the input graph has weights, PathLinker interprets them as multiplicative probabilities by default, defines the cost of a path as the product of the weights of the edges in it, and computes the most probable paths by using a log-transform. The software supports a flag to interpret edge weights as additive costs instead. If the input graph does not have weights, PathLinker can run the PageRank algorithm as a pre-processing step to generate probablistic weights from edges visitation probabilities.
See the publications referenced below for a formal description of the method.
- run.py An end-to-end implementation of the PathLinker algorithm. Given a network, a set or receptors, a set of transcription factors, and a value of k, PathLinker outputs a ranked list of edges composing the k highest scoring paths connecting any receptor to any transcription factor.
- PageRank.py An implementation of the PageRank algorithm. This algorithm can be used as a component of PathLinker, but it can also be run as a standalone tool. Given a weighted, directed network, PageRank uses the power method to compute the stationary node visitation probabilities after a teleporting random walk, i.e., a random walk with restarts.
- ksp_Astar.py An implementation of Yen's k-shortest simple paths algorithm augmented to use the A* algorithm for additional speedup. Used as a component of PathLinker, but can also be run as a standalone tool. Given a weighted, directed network, the algorithm computes the k shortest simple paths betwen a source and target node. Here a path is simple if there are no repeated nodes in the path.
Run any of the programs with the --help option for full documentation. The /example directory contains a sample usage of these programs. For all input files, lines beginning with a hash (#) are considered to be comments and ignored.
Python packages:
- NetworkX
Older releases of this code are valid in both Python 2 and Python 3. With this current version, Python 2 compatability is not guaranteed, and some unit tests fail under Python 2.
Paths of equal length may be returned and processed in a different order according to your version of Python and NetworkX. The results should be interpreted as equivalent, although the output is not textually identical.
GNU GPLv3
We will be very glad to hear from you if you use PathLinker in your work. If you publish a paper that uses PathLinker, please cite
-
Pathways on Demand: Automated Reconstruction of Human Signaling Networks. Anna Ritz, Christopher L. Poirel, Allison N. Tegge, Nicholas Sharp, Allison Powell, Kelsey Simmons, Shiv D. Kale, and T. M. Murali. npj Systems Biology and Applications, 2, Article number 16002, 2016.
-
Top-down network analysis to drive bottom-up modeling of physiological processes. Christopher L. Poirel, Richard R. Rodrigues, Katherine C. Chen, John J. Tyson, and T. M. Murali. Journal of Computational Biology, 20(5), 409-418, 2013.
We encourage experimentation with PathLinker. Please don't hesitate to contact us if you would like to collaborate!