Learning Causal Protein-Signaling Networks

Ping He, Zhi Geng, Wei Yan, Zhihai Liu – Pekin University, China

Bayesian network is widely used in causal studies. But learning causal relationships from experimental data under different conditions is a challenge. We propose an approach for structural learning from multiple data bases with external interventions. Our approach is a constraint-based method with three stages, using conditional independence tests to construct the causal network. We show the results of both simulation and real data. Finally, we discuss the advantages and disadvantages of our approach and a Bayesian approach.

NIPS 2008 workshop on causality