A systematic approach for finding the objective function and active constraints for dynamic flux balance analysis.

Dynamic flux balance analysis (DFBA) has become an instrumental modeling tool for describing the dynamic behavior of bioprocesses. DFBA involves the maximization of a biologically meaningful objective subject to kinetic constraints on the rate of consumption/production of metabolites. In this paper, we propose a systematic data-based approach for finding both the biological objective function and a minimum set of active constraints necessary for matching the model predictions to the experimental data. The proposed algorithm accounts for the errors in the experiments and eliminates the need for ad hoc choices of objective function and constraints as done in previous studies. The method is illustrated for two cases: (1) for in silico (simulated) data generated by a mathematical model for Escherichia coli and (2) for actual experimental data collected from the batch fermentation of Bordetella Pertussis (whooping cough).

Segmentation and Quantitative Analysis of Apoptosis of Chinese Hamster Ovary Cells from Fluorescence Microscopy Images.

Accurate and fast quantitative analysis of living cells from fluorescence microscopy images is useful for evaluating experimental outcomes and cell culture protocols. An algorithm is developed in this work to automatically segment and distinguish apoptotic cells from normal cells. The algorithm involves three steps consisting of two segmentation steps and a classification step. The segmentation steps are: (i) a coarse segmentation, combining a range filter with a marching square method, is used as a prefiltering step to provide the approximate positions of cells within a two-dimensional matrix used to store cells' images and the count of the number of cells for a given image; and (ii) a fine segmentation step using the Active Contours Without Edges method is applied to the boundaries of cells identified in the coarse segmentation step. Although this basic two-step approach provides accurate edges when the cells in a given image are sparsely distributed, the occurrence of clusters of cells in high cell density samples requires further processing. Hence, a novel algorithm for clusters is developed to identify the edges of cells within clusters and to approximate their morphological features. Based on the segmentation results, a support vector machine classifier that uses three morphological features: the mean value of pixel intensities in the cellular regions, the variance of pixel intensities in the vicinity of cell boundaries, and the lengths of the boundaries, is developed for distinguishing apoptotic cells from normal cells. The algorithm is shown to be efficient in terms of computational time, quantitative analysis, and differentiation accuracy, as compared with the use of the active contours method without the proposed preliminary coarse segmentation step.

A semi-empirical glycosylation model of a camelid monoclonal antibody under hypothermia cell culture conditions.

The impact of cell culture environment on the glycan distribution of a monoclonal antibody (mAb) has been investigated through a combination of experiments and modeling. A newly developed CHO DUXB cell line was cultivated at two levels of initial Glutamine (Gln) concentrations (0, 4 mM) and incubation temperatures of (33 and 37 °C) in batch operation mode. Hypothermia was applied either through the entire culture duration or only during the post-exponential phase. Beyond reducing cell growth and increasing productivity, hypothermia significantly altered the galactosylation index profiles as compared to control conditions. A novel semi-empirical dynamic model was proposed for elucidating the connections between the extracellular cell culture conditions to galactosylation index. The developed model is based on a simplified balance of nucleotides sugars and on the correlation between sugars' levels to the galactosylation index (GI). The model predictions were found to be in a good agreement with the experimental data. The proposed empirical model is expected to be useful for controlling the glycoprofiles by manipulating culture conditions.

Classification of Normal and Apoptotic Cells from Fluorescence Microscopy Images Using Generalized Polynomial Chaos and Level Set Function.

Accurate automated quantitative analysis of living cells based on fluorescence microscopy images can be very useful for fast evaluation of experimental outcomes and cell culture protocols. In this work, an algorithm is developed for fast differentiation of normal and apoptotic viable Chinese hamster ovary (CHO) cells. For effective segmentation of cell images, a stochastic segmentation algorithm is developed by combining a generalized polynomial chaos expansion with a level set function-based segmentation algorithm. This approach provides a probabilistic description of the segmented cellular regions along the boundary, from which it is possible to calculate morphological changes related to apoptosis, i.e., the curvature and length of a cell's boundary. These features are then used as inputs to a support vector machine (SVM) classifier that is trained to distinguish between normal and apoptotic viable states of CHO cell images. The use of morphological features obtained from the stochastic level set segmentation of cell images in combination with the trained SVM classifier is more efficient in terms of differentiation accuracy as compared with the original deterministic level set method.

Intrinsic fluorescence-based at situ soft sensor for monitoring monoclonal antibody aggregation.

Intrinsic fluorescence spectroscopy, in conjunction with partial least squares regression (PLSR), was investigated as a potential technique for online quality control and quantitative monitoring of Immunoglobulin G (IgG) aggregation that occurs following exposure to conditions that emulate those that can occur during protein downstream processing. Initially, the impact of three stress factors (temperature, pH, and protein concentration) on the degree of aggregation determined using size exclusion chromatography data, was investigated by performing a central composite designexperiment and applying a fitting response surface model. This investigation identified the influence of the factors as well as the operating regions with minimum propensity to induce protein aggregation. Spectral changes pertinent to the stressed samples were also investigated and found to corroborate the high sensitivity of the intrinsic fluorescence to conformational changes of the proteins under study. Ultimately, partial least squares regression was implemented to formulate two fluorescence-based soft sensors for quality control--product classification--and quantitative monitoring--concentration of monomer. The resulting regression models exhibited accurate prediction ability and good potential for in situ monitoring of monoclonal antibody downstream purification processes.

Development of a soft-sensor based on multi-wavelength fluorescence spectroscopy and a dynamic metabolic model for monitoring mammalian cell cultures.

A soft-sensor based on an Extended Kalman Filter (EKF) that combines data obtained using a fluorescence-based soft-sensor with a dynamic mechanistic model, was investigated as a tool for continuous monitoring of a Chinese hamster ovary (CHO) cell cultivation process. A standalone fluorescence based soft-sensor, which uses a combination of an empirical multivariate statistical model and measured spectra, was designed for predicting key culture variables including viable and dead cells, recombinant protein, glucose, and ammonia concentrations. The standalone fluorescence sensor was then combined with a dynamic mechanistic model within an EKF framework, for improving the prediction accuracy and generating predictions in-between sampling instances. The dynamic model used for the EKF framework was based on a structured metabolic flux analysis and mass balances. In order to calibrate the fluorescence-based empirical model and the dynamic mechanistic model, cells were grown in batch mode with different initial glucose and glutamine concentrations. To mitigate the uncertainty associated with the model structure and parameters, non-stationary disturbances were accounted for in the EKF by parameter-adaptation. It was demonstrated that the implementation of the EKF along with the dynamic model could improve the accuracy of the fluorescence-based predictions at the sampling instances. Additionally, it was shown that the major advantage of the EKF-based soft-sensor, compared to the standalone fluorescence-based counterpart, was its capability to track the temporal evolution of key process variables between measurement instances obtained by the fluorescence-based soft-sensor. This is crucial for designing control strategies of CHO cell cultures with the aim of guaranteeing product quality.

Fluorescence-based soft sensor for at situ monitoring of Chinese hamster ovary cell cultures.

Multi-wavelength fluorescence spectroscopy was investigated as a potential tool for use in monitoring key process variables that include: viable and dead cells, recombinant protein, glucose, and ammonia concentrations for Chinese hamster ovary (CHO) cells during cultivation.For the purpose of calibrating the fluorescence-based empirical model, cells were grown in batch mode with different initial glucose and glutamine concentrations.Spectrofluorometer settings were optimized to ensure reproducibility and accuracy of the acquired spectra. With the purpose of gaining qualitative insight into the evolution of the spectra, the trajectories of individual fluorophore peaks were studied during the cultivation process. Spectral changes related to biomass and secreted proteins were investigated by comparing the spectra at various stages during the downstream processing. A partial least square regression (PLSR) was used to formulate empirical models that related the input data set, i.e., the fluorescence excitation-emission matrix, to the actual state of the system including viable cell and dead cells and recombinant protein, glucose, and ammonia concentrations. The models exhibited accurate prediction ability for the process variables of interest.

Effects of nutrient levels and average culture pH on the glycosylation pattern of camelid-humanized monoclonal antibody.

The impact of operating conditions on the glycosylation pattern of humanized camelid monoclonal antibody, EG2-hFc produced by Chinese hamster ovary (CHO) cells has been evaluated by a combination of experiments and modeling. Cells were cultivated under different levels of glucose and glutamine concentrations with the goal of investigating the effect of nutrient depletion levels and ammonia build up on the cell growth and the glycoprofiles of the monoclonal antibody (Mab). The effect of average pH reduction on glycosylation level during the entire culture time or during a specific time span was also investigated. The relative abundance of glycan structures was quantified by hydrophilic interaction liquid chromatography (HILIC) and the galactosylation index (GI) and the sialylation index (SI) were determined. Lower initial concentrations of glutamine resulted in lower glucose consumption and lower cell yield but increased GI and SI levels when compared to cultures started with higher initial glutamine levels. Similarly, reducing the average pH of culture resulted in lower growth but higher SI and GI levels. These findings indicate that there is a tradeoff between cell growth, resulting Mab productivity and the achievement of desirable higher glycosylation levels. A dynamic model, based on a metabolic flux analysis (MFA), is proposed to describe the metabolism of nutrients, cell growth and Mab productivity. Finally, existing software (GLYCOVIS) that describes the glycosylation pathways was used to illustrate the impact of extracellular species on the glycoprofiles.

Dynamic metabolic modeling for a MAB bioprocess.

Production of monoclonal antibodies (MAb) for diagnostic or therapeutic applications has become an important task in the pharmaceutical industry. The efficiency of high-density reactor systems can be potentially increased by model-based design and control strategies. Therefore, a reliable kinetic model for cell metabolism is required. A systematic procedure based on metabolic modeling is used to model nutrient uptake and key product formation in a MAb bioprocess during both the growth and post-growth phases. The approach combines the key advantages of stoichiometric and kinetic models into a complete metabolic network while integrating the regulation and control of cellular activity. This modeling procedure can be easily applied to any cell line during both the cell growth and post-growth phases. Quadratic programming (QP) has been identified as a suitable method to solve the underdetermined constrained problem related to model parameter identification. The approach is illustrated for the case of murine hybridoma cells cultivated in stirred spinners.

Application of spectrofluorometry to the prediction of PHB concentrations in a fed-batch process.

On-line estimation of biopolymer production during fermentation would be a useful adjunct to the development of strategies for process control and optimization. This study examined the applicability of spectrofluorometry, along with other on-line measurements, for the prediction of poly-beta-hydroxybutyric acid (PHB) concentrations in a high-cell density fed-batch fermentation of Ralstonia eutropha. Models previously used for modelling PHB evolution with time are not sufficiently accurate for situations where transient intermediate accumulations or PHB degradation occur. Thus, the mass balance in the model was modified to account for these situations. An estimation algorithm was developed that is based on a hybrid model consisting of a dynamic mass balance of PHB where the main reaction coefficient was regressed with respect to spectrofluorometric data. The regression between the kinetic parameter and the spectrofluorometric data was accomplished using partial least squares (PLS) regression to avoid high sensitivity to noise expected from highly correlated data, such as the spectrofluorometric readings. The model accounts for dynamics of intermediates and in this way allows the prediction of dynamic behaviour in PHB concentrations that cannot be predicted with other reported mathematical models.