Experiment design (data), model selection and parameter estimation are three crucial stages of a system identification exercise. System identification of biological processes poses numerous challenges in each
stage of system identification due to inherent complexity, cost, difficulty in conducting experiments, and complexity in the model structure (high-dimensional parameter space). The objective of this work is to
address the crucial set of challenges in each stage of the identification exercise. Firstly, the precision of the parameter estimates is highly dependent upon the information contained in the data; Loss of practical
identifiability and sloppiness in the model structure are the major challenges in estimating parameters precisely and closely related to the information contained in the data. Therefore, quantifying information
is an important step in data-driven modeling. In this work, we introduce a new method for estimating information gain in the Bayesian framework using what is known as the Bhattacharyya coefficient. It is
seen that the bounds of the coefficient have an insightful interpretation naturally in terms of information gain on the parameter of interest. Secondly, assessing the goodness of model structure is a necessary task
in large-scale nonlinear system identification. Nonlinear model structures frequently suffer from loss of identifiability, anisotropic sensitivity in the parameter space, partial observability, and bifurcation. There are several analytical and numerical methods proposed to detect loss of identifiability in nonlinear model structures; however, nearly all of them suffer either from lack of scalability or generalizability to nonlinear functions. In this work, we have revisited the definition of sloppiness and used it to propose a unified framework for assessing the goodness of a model structure in terms of identifiability. Lastly, the lack of interpretability is one of the crucial challenges in black-box identification. Modeling of systems pharmacology requires interpretable models but, in a large number of cases, first principles models do
not fit the experimental data. Here we propose to attempt to integrate data-driven modeling with mechanistic modeling for better prediction and interpretability.