An empirical Bayesian approach for model-based inference of cellular signaling networks

摘要: A common challenge in systems biology is to infer mechanistic descriptions of biological process given limited observations a system. Mathematical models are frequently used represent belief about the causal relationships among proteins within signaling network. Bayesian methods provide an attractive framework for inferring validity those beliefs context available data. However, efficient sampling high-dimensional parameter space and appropriate convergence criteria barriers implementing empirical approach. The objective this study was apply Adaptive Markov chain Monte Carlo technique typical cellular pathways. As illustrative example, kinetic model early events associated with epidermal growth factor (EGF) network calibrated against dynamic measurements observed primary rat hepatocytes. criterion, based upon Gelman-Rubin potential scale reduction factor, applied predictions. posterior distributions parameters exhibited complicated structure, including significant covariance between specific broad range variance parameters. predictions, contrast, were narrowly distributed identify areas agreement collection experimental studies. In summary, approach developed confidence that one can place particular describes signal transduction mechanisms inconsistencies measurements.