Sloppy Modeling of Biological Signaling Networks
Joshua Waterfall, Cornell University
A statistical physics-inspired approach to the problem of extracting parameter values from experimental data for high dimensional, nonlinear systems such as those describing signaling networks in biological organisms is described. Correlations between parameters and a wide range of scales create a Grand Canyon appearance for the cost landscape and lead many optimization routines to be feeble. The inherent over-parameterization of the networks also motivate the study of ensembles of models as opposed to a single maximum-likelihood approach.
Abstract Author(s): Josh Waterfall