Error propagation in constraint-based modelling of Chinese hamster ovary cells
Chinese hamster ovary (CHO) cells are the cell line of choice for the production of protein biopharmaceuticals that require complex posttranslational modifications (glycosylation). Despite their success, the cell line development takes several months and is based on trial-and-error approaches (e. g. high-throughput screening and selection). Constraint-based metabolic modelling methods could potentially predict rational engineering strategies for the improvement of growth, titers and product quality. To obtain good predictions, it is essential to feed the metabolic model with highly accurate input data, especially extracellular exchange rates. In this dataset we investigated the error propagation from the extracellular metabolite concentrations to the calculated exchange rates and then to growth rate predictions by flux balance analysis (FBA) with genome-scale metabolic models of CHO. We simulated concentration profiles of 23 extracellular metabolites (glucose, lactate, ammonium, amino acids) during the exponential phase of a batch cultivation with different measurement errors of the metabolite concentrations (2-20%) and different sampling frequencies (every 6, 12 or 24 hours or at irregular sampling intervals) for 11 CHO cell lines/conditions. We identified which metabolites have the biggest impact on FBA predictions of growth rate. We also tested under what conditions we are able to reliably predict growth rate difference between two CHO cell lines/conditions. The scripts and the input/output data for the simulations of metabolite concentrations and calculation of the exchange rates are in the folder "simulated_rates". The scripts for running FBA and the input/output data are in the folder "FBA_results." The metabolic models are in the folder "models". The remaining folders correspond to the figures in the associated dataset.