A mechanistic approach for predicting mass transfer in bioreactors

A perspective-driven and technical evaluation of machine learning in bioreactor scale-up: A case-study for potential model developments

Bioreactor scale-up and scale-down have always been a topical issue for the biopharmaceutical industry and despite considerable effort, the identification of a fail-safe strategy for bioprocess development across scales remains a challenge. With the ubiquitous growth of digital transformation technologies, new scaling methods based on computer models may enable more effective scaling. This study aimed to evaluate the potential application of machine learning (ML) algorithms for bioreactor scale-up, with a specific focus on the prediction of scaling parameters. Factors critical to the development of such models were identified and data for bioreactor scale-up studies involving CHO cell-generated mAb products collated from the literature and public sources for the development of unsupervised and supervised ML models. Comparison of bioreactor performance across scales identified similarities between the different processes and primary differences between small- and large-scale bioreactors. A series of three case studies were developed to assess the relationship between cell growth and scale-sensitive bioreactor features. An embedding layer improved the capability of artificial neural network models to predict cell growth at a large-scale, as this approach captured similarities between the processes. Further models constructed to predict scaling parameters demonstrated how ML models may be applied to assist the scaling process. The development of data sets that include more characterization data with greater variability under different gassing and agitation regimes will also assist the future development of ML tools for bioreactor scaling.

An in-silico analysis of hydrodynamics and gas mass transfer characteristics in scale-down models for mammalian cell cultures

Bioprocess scale-up and technology transfer can be challenging due to multiple variables that need to be optimized during process development from laboratory scale to commercial manufacturing. Cell cultures are highly sensitive to key factors during process transfer across scales, including geometric variability in bioreactors, shear stress from impeller and sparging activity, and nutrient gradients that occur due to increasing blend times. To improve the scale-up and scale-down of these processes, it is important to fully characterize bioreactors to better understand the differences that will occur within the culture environment, especially the hydrodynamic profiles that will vary in vessel designs across scales. In this study, a comprehensive hydrodynamic characterization of the Ambr® 250 mammalian single-use bioreactor was performed using time-accurate computational fluid dynamics simulations conducted with M-Star computational fluid dynamics software, which employs lattice-Boltzmann techniques to solve the Navier-Stokes transport equations at a mesoscopic scale. The single-phase and two-phase fluid properties within this small-scale vessel were analyzed in the context of agitation hydrodynamics and mass transfer (both within the bulk fluid and the free surface) to effectively characterize and understand the differences that scale-down models possess when compared to their large-scale counterparts. The model results validate the use of computational fluid dynamics as an in-silico tool to characterize bioreactor hydrodynamics and additionally identify important free-surface transfer mechanics that need to be considered during the qualification of a scale-down model in the development of mammalian bioprocesses.

Integration Approaches to Model Bioreactor Hydrodynamics and Cellular Kinetics for Advancing Bioprocess Optimisation

Large-scale bioprocesses are increasing globally to cater to the larger market demands for biological products. As fermenter volumes increase, the efficiency of mixing decreases, and environmental gradients become more pronounced compared to smaller scales. Consequently, the cells experience gradients in process parameters, which in turn affects the efficiency and profitability of the process. Computational fluid dynamics (CFD) simulations are being widely embraced for their ability to simulate bioprocess performance, facilitate bioprocess upscaling, downsizing, and process optimisation. Recently, CFD approaches have been integrated with dynamic Cell reaction kinetic (CRK) modelling to generate valuable information about the cellular response to fluctuating hydrodynamic parameters inside large production processes. Such coupled approaches have the potential to facilitate informed decision-making in intelligent biomanufacturing, aligning with the principles of "Industry 4.0" concerning digitalisation and automation. In this review, we discuss the benefits of utilising integrated CFD-CRK models and the different approaches to integrating CFD-based bioreactor hydrodynamic models with cellular kinetic models. We also highlight the suitability of different coupling approaches for bioprocess modelling in the purview of associated computational loads.

Investigating the Hydrodynamics of Intravenous Drug Infusions

One of the most common reported adverse events for intravenous (IV) infusions are infusion site reactions, ranging from redness and pain at the site of infusion to thrombophlebitis.  The connection between drug infusion and what drives these adverse events is not well understood. To aid in understanding these phenomena, it is crucial to accurately characterize the evolving hemodynamic environment of the infusion site when developing new intravenous formulations, as too rapid dilution may cause precipitation in the vein, while too little dilution might contribute to phlebitis. In this study, a Large-Eddy Simulation (LES) turbulence modeling inside a Computational Fluid Dynamics (CFD) framework has been used to simulate the flow and mixing characteristics of an infusion entering the bloodstream. This work represents the first such study reporting transient flow fields for intravenous infusions using LES CFD simulations with a realistic non-Newtonian blood model. The output of the CFD model closely resembled the flow and mixing patterns generated in benchtop tests for infusions into a blood analogue and water as the venous fluid across a wide range of flow rates.  These models were then investigated further to compare how changes to the fluid rheology model, needle orientation and needle position within the vein resulted in altered mixing regimes at different flow rates.

CFD-based bioreactor model with proportional-integral-derivative controller functionality for dissolved oxygen and pH

A physics-based model for predicting cell culture fluid properties inside a stirred tank bioreactor with embedded PID controller logic is presented. The model evokes a time-accurate solution to the fluid velocity field and overall volumetric mass transfer coefficient, as well as the ongoing effects of interfacial mass transfer, species mixing, and aqueous chemical reactions. The modeled system also includes a direct coupling between process variables and system control variables via embedded controller logic. Satisfactory agreement is realized between the model prediction and measured bioreactor data in terms of the steady-state operating conditions and the response to setpoint changes. Simulation runtimes are suitable for industrial research and design timescales.

Experimental and computational characterization of mass transfer in high turndown bioreactors

Single-use bioreactors (SUBs, or disposable bioreactors) are extensively used for the clinical and commercial production of biologics. Despite widespread application, minimal results have been reported utilizing the turndown ratio; an operation mode where the working range of the bioreactor can be expanded to include low fluid volumes. In this work, a systematic investigation into free surface mass transfer and cell growth in high turndown single-use bioreactors is presented. This approach, which combines experimental mass transfer measurements with numerical simulation, deconvolutes the combined effects of headspace mixing and the free surface convective mass transfer on cell growth. Under optimized conditions, mass transfer across the interface alone may be sufficient to satisfy oxygen demands of the cell culture. Within the context of high turndown bioreactors, this finding provides a counterpoint to traditional sparge-based bioreactor operational philosophy. Multiple monoclonal antibody-producing cell lines grown using this high turndown approach showed similar viable cell densities to those cells expanded using a traditional cell bag rocker. Furthermore, cells taken directly from the turndown expansion and placed into production showed identical growth characteristics to traditionally expanded cultures. Taken together, these results suggest that the Xcellerex SUB can be run at a 5:1 working volume as a seed to itself, with no need for system modifications, potentially simplifying preculture operations.

Transferring Bubble Breakage Models Tailored for Euler-Euler Approaches to Euler-Lagrange Simulations

Most bubble breakage models have been developed for multiphase simulations using Euler-Euler (EE) approaches. Commonly, they are linked with population balance models (PBM) and are validated by making use of Reynolds-averaged Navier-Stokes (RANS) turbulence models. The latter, however, may be replaced by alternate approaches such as Large Eddy simulations (LES) that play a pivotal role in current developments based on lattice Boltzmann (LBM) technologies. Consequently, this study investigates the possibility of transferring promising bubble breakage models from the EE framework into Euler-Lagrange (EL) settings aiming to perform LES. Using our own model, it was possible to reproduce similar bubble size distributions (BSDs) for EL and EE simulations. Therefore, the critical Weber (Wecrit) number served as a threshold value for the occurrence of bubble breakage events. Wecrit depended on the bubble daughter size distribution (DSD) and a set minimum time between two consecutive bubble breakage events. The commercial frameworks Ansys Fluent and M-Star were applied for EE and EL simulations, respectively. The latter enabled the implementation of LES, i.e., the use of a turbulence model with non-time averaged entities. By properly choosing Wecrit, it was possible to successfully transfer two commonly applied bubble breakage models from EE to EL. Based on the mechanism of bubble breakage, Wecrit values of 7 and 11 were determined, respectively. Optimum Wecrit were identified as fitting the shape of DSDs, as this turned out to be a key criterion for reaching optimum prediction quality. Optimum Wecrit values hold true for commonly applied operational conditions in aerated bioreactors, considering water as the matrix.

Predictive Drug Release Modeling Across Dissolution Apparatuses I and II using Computational Fluid Dynamics

A modeling process is developed and validated with which active pharmaceutical ingredient (API) release is predicted across the United States Pharmacopeia (USP) dissolution apparatuses I and II based on limited experimental dissolution data (at minimum two dissolution profiles at different apparatus settings). The process accounts for formulation-specific drug release behavior and hydrodynamics in the apparatuses over the range of typical agitation rates and medium volumes. This modeling process involves measurement of experimental mass transfer coefficients via a conventional mass balance and the relationship of said mass transfer coefficients to hydrodynamics and apparatus setting via computational fluid dynamics (CFD). A novel 1-D model is hence established, which provided calibration data for a particular formulation, can model mass transfer coefficients and their corresponding drug release at apparatus configurations of interest. Based on validation against experimental data produced from five erosion-based formulations over a range of apparatus configurations, accuracy within 8 %LA (labelled amount of API) and an average root mean square deviation of 3 %LA is achieved. With this predictive capability, minimizing the number of dissolution experiments and the amount of chemical materials needed during method development appears feasible.

CFD supported scale up of perfusion bioreactors in biopharma

The robust scale up of perfusion systems requires comparable conditions over all scales to ensure equivalent cell culture performance. As cells in continuous processes circulate outside the bioreactor, performance losses may arise if jet flow and stirring cause a direct connection between perfusion feed and return. Computational fluid dynamics can be used to identify such short circuit flows, assess mixing efficiencies, and eventually adapt the perfusion setup. This study investigates the scale up from a 2 L glass bioreactor to 100 L and 500 L disposable pilot scale systems. Highly resolved Lattice Boltzmann Large Eddy simulations were performed in single phase and mixing efficiencies (Emix) furthermore experimentally validated in the 2 L system. This evaluation gives insight into the flow pattern, the mixing behavior and information on cell residence time inside the bioreactors. No geometric adaptations in the pilot scale systems were necessary as Emix was greater than 90% for all conditions tested. Two different setups were evaluated in 2 L scale where the direction of flow was changed, yielding a difference in mixing efficiency of 10%. Nevertheless, since Emix was confirmed to be >90% also for both 2 L setups and the determined mixing times were in a similar range for all scales, the 2 L system was deemed to be a suitable scale down model. The results demonstrate how computational fluid dynamic models can be used for rational process design of intensified production processes in the biopharmaceutical industry.

Refolding in the modern biopharmaceutical industry

Inclusion bodies (IBs) often emerge upon overexpression of recombinant proteins in E. coli. From IBs, refolding is necessary to generate the native protein that can be further purified to obtain pure and active biologicals. This work focusses on refolding as a significant process step during biopharmaceutical manufacturing with an industrial perspective. A theoretical and historical background on protein refolding gives the reader a starting point for further insights into industrial process development. Quality requirements on IBs as starting material for refolding are discussed and further economic and ecological aspects are considered with regards to buffer systems and refolding conditions. A process development roadmap shows the development of a refolding process starting from first exploratory screening rounds to scale-up and implementation in manufacturing plant. Different aspects, with a direct influence on yield, such as the selection of chemicals including pH, ionic strength, additives, etc., and other often neglected aspects, important during scale-up, such as mixing, and gas-fluid interaction, are highlighted with the use of a quality by design (QbD) approach. The benefits of simulation sciences (process simulation and computer fluid dynamics) and process analytical technology (PAT) for seamless process development are emphasized. The work concludes with an outlook on future applications of refolding and highlights open research inquiries.

Modeling multiphase fluid flow, mass transfer, and chemical reactions in bioreactors using large-eddy simulation

We present a transient large eddy simulation (LES) modeling approach for simulating the interlinked physics describing free surface hydrodynamics, multiphase mixing, reaction kinetics, and mass transport in bioreactor systems. Presented case-studies include non-reacting and reacting bioreactor systems, modeled through the inclusion of uniform reaction rates and more complex biochemical reactions described using Contois type kinetics. It is shown that the presence of reactions can result in a non-uniform spatially varying species concentration field, the magnitude and extent of which is directly related to the reaction rates and the underlying variations in the local volumetric mass transfer coefficient.

Engineering Considerations to Produce Bioactive Compounds from Plant Cell Suspension Culture in Bioreactors

The large-scale production of plant-derived secondary metabolites (PDSM) in bioreactors to meet the increasing demand for bioactive compounds for the treatment and prevention of degenerative diseases is nowadays considered an engineering challenge due to the large number of operational factors that need to be considered during their design and scale-up. The plant cell suspension culture (CSC) has presented numerous benefits over other technologies, such as the conventional whole-plant extraction, not only for avoiding the overexploitation of plant species, but also for achieving better yields and having excellent scaling-up attributes. The selection of the bioreactor configuration depends on intrinsic cell culture properties and engineering considerations related to the effect of operating conditions on thermodynamics, kinetics, and transport phenomena, which together are essential for accomplishing the large-scale production of PDSM. To this end, this review, firstly, provides a comprehensive appraisement of PDSM, essentially those with demonstrated importance and utilization in pharmaceutical industries. Then, special attention is given to PDSM obtained out of CSC. Finally, engineering aspects related to the bioreactor configuration for CSC stating the effect of the operating conditions on kinetics and transport phenomena and, hence, on the cell viability and production of PDSM are presented accordingly. The engineering analysis of the reviewed bioreactor configurations for CSC will pave the way for future research focused on their scaling up, to produce high value-added PDSM.

Towards enhancement of gas-liquid mass transfer in bioelectrochemical systems: Validation of a robust CFD model

Mass transfer has been identified as a major bottleneck in gas fermentation and microbial conversion of carbon dioxide to chemicals. We present a pragmatic and validated Computational Fluid Dynamics (CFD) model for mass transfer in bioelectrochemical systems. Experiments were conducted to measure mixing times and mass transfer in a Duran bottle and an H-cell. An Eulerian-Eulerian framework with a simplified model for the bubble size distribution (BSD) was developed that utilized only one additional equation for the bubble number density while including the breakup and coalescence. Validations of the CFD model for mixing times showed that the predictions were within the confidence intervals of the measurements, verifying the model's capability in simulating the hydrodynamics. Further validations were performed using constant and varying bubble diameters for the mass transfer. The results showed the benefits of a simplified BSD model, as it yielded improvements of seven and four times in accuracy when assessed against the experimental data for the Duran bottle and H-cell, respectively. Modeling of the H-cell predicted that a lower stirring rate improves mass transfer compared with higher stirring rates, which is of great importance when designing microbial cultivation processes. The model offers a feasible framework for advanced modeling of gas fermentation and microbial electrosynthesis.

Validation of Novel Lattice Boltzmann Large Eddy Simulations (LB LES) for Equipment Characterization in Biopharma

Detailed process and equipment knowledge is crucial for the successful production of biopharmaceuticals. An essential part is the characterization of equipment for which Computational Fluid Dynamics (CFD) is an important tool. While the steady, Reynolds-averaged Navier–Stokes (RANS) k − ε approach has been extensively reviewed in the literature and may be used for fast equipment characterization in terms of power number determination, transient schemes have to be further investigated and validated to gain more detailed insights into flow patterns because they are the method of choice for mixing time simulations. Due to the availability of commercial solvers, such as M-Star CFD, Lattice Boltzmann simulations have recently become popular in the industry, as they are easy to set up and require relatively low computing power. However, extensive validation studies for transient Lattice Boltzmann Large Eddy Simulations (LB LES) are still missing. In this study, transient LB LES were applied to simulate a 3 L bioreactor system. The results were compared to novel 4D particle tracking (4D PTV) experiments, which resolve the motion of thousands of passive tracer particles on their journey through the bioreactor. Steady simulations for the determination of the power number followed a structured workflow, including grid studies and rotating reference frame volume studies, resulting in high prediction accuracy with less than 11% deviation, compared to experimental data. Likewise, deviations for the transient simulations were less than 10% after computational demand was reduced as a result of prior grid studies. The time averaged flow fields from LB LES were in good accordance with the novel 4D PTV data. Moreover, 4D PTV data enabled the validation of transient flow structures by analyzing Lagrangian particle trajectories. This enables a more detailed determination of mixing times and mass transfer as well as local exposure times of local velocity and shear stress peaks. For the purpose of standardization of common industry CFD models, steady RANS simulations for the 3 L vessel were included in this study as well.