|
 |
|||||||||||||||||||
|
|||||||||||||||||||
|
|||||||||||||||||||
 |
 |
|
|
a Banting & Best Department of Medical Research, Donnelly Centre for Cellular and Biomolecular Research, University of Toronto, Toronto, Ontario M5S 3E1, Canada b Department of Medical Genetics and Microbiology, University of Toronto, Toronto, Ontario M5S 3E1, Canada
Key Words: reverse engineering • perturbation networks • alternative pathway approach • pathnet
Address for correspondence: Yunchen Gong, Banting & Best Department of Medical Research, Donnelly Center for Cellular and Biomolecular Research, University of Toronto, 160 College Street, Ontario M5S 3E1, Canada. Voice: (416) 946-0905; fax: (416) 978-8287. yunchen.gong{at}utoronto.ca Zhaolei.Zhang{at}utoronto.ca
Cell perturbation data are a very important resource to analyze and reconstruct cell-signaling networks. To facilitate the utilization of this type of data and enable large-scale and automated network reconstruction effort, we have already developed a data structure for storing cell perturbation results and deducing perturbation networks (CellFrame). For automating network analysis, we here propose a computational method called the "alternative pathway approach" (ALPA) in this work. This method can validate the signaling networks with conditional perturbation data extracted from published experiments, and can suggest additional tests to improve the network. It searches the alternative pathway space between all pairs of nodes, constructs pathnets (set of pathways between two nodes) and validates the network edges using conditional perturbation. For pathnets without conditional data, experiments with the fewest number of perturbations or the most parsimonious are designed. For pathnets without experimentally derived or detected pair-wise interactions, ALPA can predict the potential effects or propose additional pathways to expand the existing network. We have tested the ALPA method on the TNF
 -MAPK signaling cascade; the reconstructed network is consistent with the consensus model. We also used the ALPA algorithm to analyze a yeast gene perturbation network, using the data from the Rosetta compendium of expression profiles, which demonstrates that it can be also used for large-scale analysis. We compared the performance of the ALPA approach with Boolean and Bayesian network algorithms for their efficiency and accuracy, respectively, in network construction.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
