2019/06/24 by Nibodh Boddupalli, Aqib Hasnain, Boddupalli, Nibodh +5
Biochemistry, Genetics and Molecular Biology · Physics and Astronomy · #Dynamical Systems (math.DS) #FOS: Biological sciences #FOS: Electrical engineering #FOS: Mathematics #Gene Regulatory Network Analysis #Model Reduction and Neural Networks #Molecular Networks (q-bio.MN) #Optimization and Control (math.OC) #Protein Structure and Dynamics #Systems and Control (eess.SY) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1906.10274
openalex publication_date 2019/06/24 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
It is hard to identify nonlinear biological models strictly from data, with\nresults that are often sensitive to experimental conditions. Automated\nexperimental workflows and liquid handling enables unprecedented throughput, as\nwell as the capacity to generate extremely large datasets. We seek to develop\ngeneralized identifiability conditions for informing the design of automated\nexperiments to discover predictive nonlinear biological models. For linear\nsystems, identifiability is characterized by persistence of excitation\nconditions. For nonlinear systems, no such persistence of excitation conditions\nexist. We use the input-Koopman operator method to model nonlinear systems and\nderive identifiability conditions for open-loop systems initialized from a\nsingle initial condition. We show that nonlinear identifiability is\nintrinsically tied to the rank of a given dataset's power spectral density,\ntransformed through the lifted Koopman observable space. We illustrate these\nidentifiability conditions with a simulated synthetic gene circuit model, the\nrepressilator. We illustrate how rank degeneracy in datasets results in\noverfitted nonlinear models of the repressilator, resulting in poor predictive\naccuracy. Our findings provide novel experimental design criteria for discovery\nof globally predictive nonlinear models of biological phenomena.\n