Generating Executable Models From Signaling Network Connectivity and Semi-Quantitative Proteomic Measurements

Derek Ruths*, Luay Nakhleh

School of Computer Science, McGill University, Quebec, Montreal Canada. derek.ruths@cs.mcgill.ca

Proc LSS Comput Syst Bioinform Conf. August, 2010. Vol. 9, p. 136-145. Full-Text PDF

*To whom correspondence should be addressed.

Executable biology is a discipline that is concerned with turning the specifications of a biological system into a computational model that can be simulated under different conditions to produce predictions about the behavior of the system. In this paper, we propose a computational framework consisting of a generalized execution strategy for signaling networks as well as a method for learning executable models from connectivity-maps and proteomic data pertaining to a specific signaling network. We call these data sources semi-quantitative because often they characterize the behavior of the system without providing reliably exact numerical measurements. To the best of our knowledge this is the first use of semi-quantitative data for building predictive models of biochemical systems. Using our framework, we generate an executable model of a network of signaling pathways downstream of the epidermal growth factor receptor (EGFR) in the MCF-7 cell line. Using this executable model, we determine that our method performs as well as existing methods while using orders of magnitude less training data to achieve a comparable degree of accuracy.