(262d) Specifying Experiment Design Points to Discern Between Hypotheses for Models of Intracellular Signaling Pathways Using Sparse Grids

A quantitative sparse grid-based experiment design algorithm has been developed for intracellular signaling models that augments traditional statistical approaches for the design of experiments. The algorithm selects the minimal number of experimental measurements (design points) that will discriminate between hypotheses for given experimental conditions. System hypotheses can be encoded in parameter values and model structures. For most systems, the initial available experimental data is incomplete, sparse and noisy, and parameters can be highly correlated and vary greatly in sensitivities. Therefore, different sets of parameter values can produce model simulations that fit the data, while leading to vastly different predictions in the dynamics of unmeasured elements. The proposed algorithm identifies and characterizes acceptable parameter sets, those that support model simulations that adequately fit the available data, using focused sparse grid-based interpolation and clustering algorithms. A greedy algorithm produces a minimal experiment design to discriminate between competing hypotheses. This approach differs from common methods utilizing the Fisher Information Matrix, as it is a global method, does not approximate or linearize the model, and is still computationally efficient. The experiment design algorithm is demonstrated on a mitogen activated protein kinase cascade model. The results show that even with an established model structure, the system dynamics are still highly uncertain when experimental data is limited. Nevertheless, the algorithm suggests additional experimental data points that enable the successful, simultaneous discrimination between possible model structures and acceptable parameter values. A global sensitivity analysis was used to evaluate the effectiveness of the focusing algorithm to identify the acceptable parameter regions. The correlation of the global sensitivity analysis rank and the degree of variation in the representative parameter values can indicate whether the parameter space was sufficiently sampled to resolve the hypotheses encoded through the uncertain parameter values and model structures. This sparse grid-based experiment design process provides a systematic and efficient exploration over the entire uncertain parameter space of potential model structures to resolve the uncertainty in the intracellular signaling model dynamics.