Dynamic metabolic modeling based on in vivo, time series measurements
Event Type
Student Poster
TimeTuesday, July 306:30pm - 8:30pm
LocationCrystal Foyer and Crystal B
DescriptionMultiple modeling approaches of metabolic networks have been developed for Neurospora crassa. However, common methods often assume steady state and fixed network topologies, such as Flux Balance Analysis (FBA). In order to further analyse metabolism, dynamic modeling with network topology identification is needed. Identifying kinetic parameters and topology has been difficult task. Metabolic time-series measurements can provide crucial constraints on these parameters, but the conventional approach is labour-intensive and subject to biological and technical variability. In this project, using continuous in vivo monitoring of metabolism by NMR (CIVM-NMR), we obtained time-series metabolic measurements on single N. crassa cultures, which eliminates extraction variance. Based on these time-series measurements, we built a model for the central metabolic pathway (~50 reactions), including, inter alia, glycolysis, fermentation and TCA cycle. Reactions in the network are modeled by ODE (Ordinary Differential Equation). A Metropolis–Hastings based MCMC (Markov chain Monte Carlo) optimization process is implemented to explore the parameter space and find a large sample of models consistent with the measurements. Evaluation of existing allosteric regulation is tested by fixed topology models, and new regulation will be estimated by variable topology models. Different regulation functions (power equation and Hill equation) are compared for numerical performance and fitting accuracy. Parallel tempering will also be used to speed-up optimization. ODE simulation is CPU intensive and takes most of the time in optimization. Parameter space exploration by MCMC takes tens thousands of sweeps. MPI resources from GACRC and XSEDE are currently used for model identification in this project.