(680a) DATA-Driven Reconstruction Of Biological Networks Using A Nonlinear MODEL Formulation

Reconstruction of biological networks is of great interest for learning from large volume of experimental data and facilitating interpretation in biology. There are several methods developed for data-driven reconstruction of biological networks. However, to the best of our knowledge, only a limited number of algorithms have been designed for systematic network reconstruction to account for the nonlinearity of biological systems. Here, we have introduced a novel method to reconstruct nonlinear biological networks. In this method, we use a quadratic nonlinear model as the representation of second-order Taylor series expansion of a nonlinear system around an arbitrary point of interest. As a heuristic, we use a threshold on the correlation between the input/output pairs to identify the informative (significant) inputs in the dataset. We apply LASSO on the significant inputs to shrink some of the small coefficients to zero. Simulation results show considerable improvements in predicting the response of the system and fair improvement in accuracy and sensitivity of the network identified. We have used both synthetic and real biological systems to test our methodology. For the synthetic system, we use different types of nonlinearities to validate the generality of the approach. In the case of real system, we have applied this algorithm to a phosphoprotein (input)/cytokine (output) dataset in RAW 264.7 macrophage cells for which the proposed nonlinear formulation achieves a better performance than a linear model-based formulation of the model.