(601e) Causal Inference of Time-Varying Signaling Networks Using Dynamic Data

Intracellular signaling pathways transmit extracellular signals to the nucleus to regulate gene expression resulting in various cellular functions including adaptation to environmental changes. These signaling pathways are dynamic and nonlinear in nature. They can be modeled as networks. To capture the dynamical nature of such networks, we have developed a novel framework to reconstruct the temporally evolving networks using dynamic data [1]. We use a vector autoregressive model formulation to capture the notion of Granger causality. We perform statistical significance testing on the estimated coefficients of the model to identify potentially causal connections in the network. We have applied the approach to a large-scale dynamic dataset on phosphoproteomic measurements in RAW 264.7 macrophage cells. Proteins generally become active when phosphorylated. Thus, phosphoproteomic measurements serve as readouts for the activation of intracellular signaling pathways. Using this dynamic data, we have developed a three-stage causal network capturing the evolution of the system from one stage to another. Stage 1 of the network mainly consists of edges active during early response to various stimuli/ligands whereas stages 2 and 3 mostly capture the fully active state of the signaling pathways.

 Reference:

[1] Masnadi-Shirazi, M, Maurya, MR, Subramaniam,S. Time-varying causal inference from phosphoproteomic measurements in macrophage cells. IEEE Trans Biomed Circuits Syst 2014; 8(1):74-86.