Challenges in the Development of Soft Sensors for Bioprocesses: A Critical Review

A Hybrid Soft Sensor Approach Combining Partial Least-Squares Regression and an Unscented Kalman Filter for State Estimation in Bioprocesses

Real-time information on key state variables during fermentation is crucial for the effective optimization and control of bioprocesses. Specialized sensors for online or at-line monitoring of these variables are often associated with high costs, especially during early-stage process optimization. In this study, fed-batch processes of an L-phenylalanine (L-phe) production process were carried out using a recombinant Escherichia coli strain under varying inducer concentrations. The available online process variables from the L-phe production process were used to estimate the state variables biomass, glycerol, L-phe, acetate, and L-tyrosine (L-tyr) via partial least-squares regression (PLSR). These predictions were then incorporated as measurements into an unscented Kalman filter (UKF). The filter uses a coarse-grained model as a state estimator, which, in addition to extracellular variables, also provides information on intracellular states. The results of PLSR showed very good prediction accuracy for L-phe, moderate accuracy for glycerol, biomass, and L-tyr and poor performance for acetate concentrations. In combination with the UKF, the estimation of the L-phe concentrations was greatly improved compared to the CGM, whereas further improvement is still needed for the remaining state variables.

An Analysis of Monitoring Solutions for CAR T Cell Production

The chimeric antigen receptor T cell (CAR T) therapy has shown remarkable results in treating certain cancers. It involves genetically modifying a patient's T cells to recognize and attack cancer cells. Despite its potential, CAR T cell therapy is complex and costly and requires the integration of multiple technologies and specialized equipment. Further research is needed to achieve the maximum potential of CAR T cell therapies and to develop effective and efficient methods for their production. This paper presents an overview of current measurement methods used in the key steps of the production of CAR T cells. The study aims to assess the state of the art in monitoring solutions and identify their potential for online monitoring. The results of this paper contribute to the understanding of measurement methods in CAR T cell manufacturing and identify areas where on-line monitoring can be improved. Thus, this research facilitates progress toward the development of effective monitoring of CAR T cell therapies.

Series-connected data-based model to estimate effluent chemical oxygen demand in industrial wastewater treatment process

Continuous monitoring of chemical oxygen demand (COD) is essential to ensure efficient and sustainable wastewater treatment and regulatory compliance. However, traditional hardware measurements are laborious, infrequent and costly. In this research, a cost-effective real-time alternative is presented. ARMAX, ARX, Neural Network and PLSR model structures were identified and tested for finding data-based model for real-time estimation of effluent COD in a full-scale industrial WWTP. To aim for estimating effluent COD without physically measuring it, a novel chain of two estimators was created by connecting in series the identified influent and effluent COD models. A comprehensive and systematic model identification was carried out to find the model inputs, delays, parameters and training windows using an exhaustive search algorithm. The results showed that using solely linear model structure it is possible to identify sufficiently accurate (R: 0.67, MAPE: 7.33%) and practical (interpretable and implementable) data-based estimation model which has predictive ability even up to 20 h horizon. As the series-connected model structure reaches the required margin of error it has potential for real-world industrial usage alongside or even replacing the hardware online sensor. Estimation model enhances resiliency and provides real-time insights into effluent quality in varying operating conditions and during unexpected disturbances.

Continuous Process Verification 4.0 application in upstream: adaptiveness implementation managed by AI in the hypoxic bioprocess of the Pichia pastoris cell factory

The experimental approach developed in this research demonstrated how the cloud, the Internet of Things (IoT), edge computing, and Artificial Intelligence (AI), considered key technologies in Industry 4.0, provide the expected horizon for adaptive vision in Continued Process Verification (CPV), the final stage of Process Validation (PV). Pichia pastoris producing Candida rugosa lipase 1 under the regulation of the constitutive GAP promoter was selected as an experimental bioprocess. The bioprocess worked under hypoxic conditions in carbon-limited fed-batch cultures through a physiological control based on the respiratory quotient (RQ). In this novel bioprocess, a digital twin (DT) was built and successfully tested. The implementation of online sensors worked as a bridge between the microorganism and AI models, to provide predictions from the edge and the cloud. AI models emulated the metabolism of Pichia based on critical process parameters and actionable factors to achieve the expected quality attributes. This innovative AI-aided Adaptive-Proportional Control strategy (AI-APC) improved the reproducibility comparing to a Manual-Heuristic Control strategy (MHC), showing better performance than the Boolean-Logic-Controller (BLC) tested. The accuracy, indicated by the Mean Relative Error (MRE), was for the AI-APC lower than 4%, better than the obtained for MHC (10%) and BLC (5%). Moreover, in terms of precision, the same trend was observed when comparing the Root Mean Square Deviation (RMSD) values, becoming lower as the complexity of the controller increases. The successful automatic real time control of the bioprocess orchestrated by AI models proved the 4.0 capabilities brought by the adaptive concept and its validity in biopharmaceutical upstream operations.

Advances and opportunities in process analytical technologies for viral vector manufacturing

Viral vectors are an emerging, exciting class of biologics whose application in vaccines, oncology, and gene therapy has grown exponentially in recent years. Following first regulatory approval, this class of therapeutics has been vigorously pursued to treat monogenic disorders including orphan diseases, entering hundreds of new products into pipelines. Viral vector manufacturing supporting clinical efforts has spurred the introduction of a broad swath of analytical techniques dedicated to assessing the diverse and evolving panel of Critical Quality Attributes (CQAs) of these products. Herein, we provide an overview of the current state of analytics enabling measurement of CQAs such as capsid and vector identities, product titer, transduction efficiency, impurity clearance etc. We highlight orthogonal methods and discuss the advantages and limitations of these techniques while evaluating their adaptation as process analytical technologies. Finally, we identify gaps and propose opportunities in enabling existing technologies for real-time monitoring from hardware, software, and data analysis viewpoints for technology development within viral vector biomanufacturing.

Soft bioreactor systems: a necessary step toward engineered MSK soft tissue?

A key objective of tissue engineering (TE) is to produce in vitro funcional grafts that can replace damaged tissues or organs in patients. TE uses bioreactors, which are controlled environments, allowing the application of physical and biochemical cues to relevant cells growing in biomaterials. For soft musculoskeletal (MSK) tissues such as tendons, ligaments and cartilage, it is now well established that applied mechanical stresses can be incorporated into those bioreactor systems to support tissue growth and maturation via activation of mechanotransduction pathways. However, mechanical stresses applied in the laboratory are often oversimplified compared to those found physiologically and may be a factor in the slow progression of engineered MSK grafts towards the clinic. In recent years, an increasing number of studies have focused on the application of complex loading conditions, applying stresses of different types and direction on tissue constructs, in order to better mimic the cellular environment experienced in vivo. Such studies have highlighted the need to improve upon traditional rigid bioreactors, which are often limited to uniaxial loading, to apply physiologically relevant multiaxial stresses and elucidate their influence on tissue maturation. To address this need, soft bioreactors have emerged. They employ one or more soft components, such as flexible soft chambers that can twist and bend with actuation, soft compliant actuators that can bend with the construct, and soft sensors which record measurements in situ. This review examines types of traditional rigid bioreactors and their shortcomings, and highlights recent advances of soft bioreactors in MSK TE. Challenges and future applications of such systems are discussed, drawing attention to the exciting prospect of these platforms and their ability to aid development of functional soft tissue engineered grafts.

Collaborative Multiple Players to Address Label Sparsity in Quality Prediction of Batch Processes

For decades, soft sensors have been extensively renowned for their efficiency in real-time tracking of expensive variables for advanced process control. However, despite the diverse efforts lavished on enhancing their models, the issue of label sparsity when modeling the soft sensors has always posed challenges across various processes. In this paper, a fledgling technique, called co-training, is studied for leveraging only a small ratio of labeled data, to hone and formulate a more advantageous framework in soft sensor modeling. Dissimilar to the conventional routine where only two players are employed, we investigate the efficient number of players in batch processes, making a multiple-player learning scheme to assuage the sparsity issue. Meanwhile, a sliding window spanning across both time and batch direction is used to aggregate the samples for prediction, and account for the unique 2D correlations among the general batch process data. Altogether, the forged framework can outperform the other prevalent methods, especially when the ratio of unlabeled data is climbing up, and two case studies are showcased to demonstrate its effectiveness.

Synergic kinetic and physiological control to improve the efficiency of Komagataella phaffii recombinant protein production bioprocesses

The yeast Komagataella phaffii (Pichia pastoris) is currently considered a versatile and highly efficient host for recombinant protein production (RPP). Interestingly, the regulated application of specific stress factors as part of bioprocess engineering strategies has proven potential for increasing the production of recombinant products. This study aims to evaluate the impact of controlled oxygen-limiting conditions on the performance of K. phaffii bioprocesses for RPP in combination with the specific growth rate (μ) in fed-batch cultivations. In this work, Candida rugosa lipase 1 (Crl1) production, regulated by the constitutive GAP promoter, growing at different nominal μ (0.030, 0.065, 0.100 and 0.120 h-1 ) under both normoxic and hypoxic conditions in carbon-limiting fed-batch cultures is analysed. Hypoxic fermentations were controlled at a target respiratory quotient (RQ) of 1.4, with excellent performance, using an innovative automated control based on the stirring rate as the manipulated variable developed during this study. The results conclude that oxygen limitation positively affects bioprocess efficiency under all growing conditions compared. The shift from respiratory to respiro-fermentative metabolism increases bioprocess productivity by up to twofold for the specific growth rates evaluated. Moreover, the specific product generation rate (qp ) increases linearly with μ, regardless of oxygen availability. Furthermore, this hypoxic boosting effect was also observed in the production of Candida antarctica lipase B (CalB) and pro-Rhizopus oryzae lipase (proRol), thus proving the synergic effect of kinetic and physiological stress control. Finally, the Crl1 production scale-up was conducted successfully, confirming the strategy's scalability and the robustness of the results obtained at the bench-scale level.

Current challenges and recent advances on the path towards continuous biomanufacturing

Continuous biopharmaceutical manufacturing is currently a field of intense research due to its potential to make the entire production process more optimal for the modern, ever-evolving biopharmaceutical market. Compared to traditional batch manufacturing, continuous bioprocessing is more efficient, adjustable, and sustainable and has reduced capital costs. However, despite its clear advantages, continuous bioprocessing is yet to be widely adopted in commercial manufacturing. This article provides an overview of the technological roadblocks for extensive adoptions and points out the recent advances that could help overcome them. In total, three key areas for improvement are identified: Quality by Design (QbD) implementation, integration of upstream and downstream technologies, and data and knowledge management. First, the challenges to QbD implementation are explored. Specifically, process control, process analytical technology (PAT), critical process parameter (CPP) identification, and mathematical models for bioprocess control and design are recognized as crucial for successful QbD realizations. Next, the difficulties of end-to-end process integration are examined, with a particular emphasis on downstream processing. Finally, the problem of data and knowledge management and its potential solutions are outlined where ontologies and data standards are pointed out as key drivers of progress.

Soft Sensing of LPG Processes Using Deep Learning

This study investigates the integration of soft sensors and deep learning in the oil-refinery industry to improve monitoring efficiency and predictive accuracy in complex industrial processes, particularly de-ethanization and debutanization. Soft sensor models were developed to estimate critical variables such as the C2 and C5 contents in liquefied petroleum gas (LPG) after distillation and the energy consumption of distillation columns. The refinery's LPG purification process relies on periodic sampling and laboratory analysis to maintain product specifications. The models were tested using data from actual refinery operations, addressing challenges such as scalability and handling dirty data. Two deep learning models, an artificial neural network (ANN) soft sensor model and an ensemble random forest regressor (RFR) model, were developed. This study emphasizes model interpretability and the potential for real-time updating or online learning. The study also proposes a comprehensive, iterative solution for predicting and optimizing component concentrations within a dual-column distillation system, highlighting its high applicability and potential for replication in similar industrial scenarios.

Generalizability of Soft Sensors for Bioprocesses through Similarity Analysis and Phase-Dependent Recalibration

A soft sensor concept is typically developed and calibrated for individual bioprocesses in a time-consuming manual procedure. Following that, the prediction performance of these soft sensors degrades over time, due to changes in raw materials, biological variability, and modified process strategies. Through automatic adaptation and recalibration, adaptive soft sensor concepts have the potential to generalize soft sensor principles and make them applicable across bioprocesses. In this study, a new generalized adaptation algorithm for soft sensors is developed to provide phase-dependent recalibration of soft sensors based on multiway principal component analysis, a similarity analysis, and robust, generalist phase detection in multiphase bioprocesses. This generalist soft sensor concept was evaluated in two multiphase bioprocesses with various target values, media, and microorganisms. Consequently, the soft sensor concept was tested for biomass prediction in a Pichia pastoris process, and biomass and protein prediction in a Bacillus subtilis process, where the process characteristics (cultivation media and cultivation strategy) were varied. High prediction performance was demonstrated for P. pastoris processes (relative error = 6.9%) as well as B. subtilis processes in two different media during batch and fed-batch phases (relative errors in optimized high-performance medium: biomass prediction = 12.2%, protein prediction = 7.2%; relative errors in standard medium: biomass prediction = 12.8%, protein prediction = 8.8%).

Assessing the capabilities of 2D fluorescence monitoring in microtiter plates with data-driven modeling for secondary substrate limitation experiments of Hansenula polymorpha

BACKGROUND

Non-invasive online fluorescence monitoring in high-throughput microbioreactors is a well-established method to accelerate early-stage bioprocess development. Recently, single-wavelength fluorescence monitoring in microtiter plates was extended to measurements of highly resolved 2D fluorescence spectra, by introducing charge-coupled device (CCD) detectors. Although introductory experiments demonstrated a high potential of the new monitoring technology, an assessment of the capabilities and limits for practical applications is yet to be provided.

RESULTS

In this study, three experimental sets introducing secondary substrate limitations of magnesium, potassium, and phosphate to cultivations of a GFP-expressing H. polymorpha strain were conducted. This increased the complexity of the spectral dynamics, which were determined by 2D fluorescence measurements. The metabolic responses upon growth limiting conditions were assessed by monitoring of the oxygen transfer rate and extensive offline sampling. Using only the spectral data, subsequently, partial least-square (PLS) regression models for the key parameters of glycerol, cell dry weight, and pH value were generated. For model calibration, spectral data of only two cultivation conditions were combined with sparse offline sampling data. Applying the models to spectral data of six cultures not used for calibration, resulted in an average relative root-mean-square error (RMSE) of prediction between 6.8 and 6.0%. Thus, while demanding only sparse offline data, the models allowed the estimation of biomass accumulation and glycerol consumption, even in the presence of more or less pronounced secondary substrate limitation.

CONCLUSION

For the secondary substrate limitation experiments of this study, the generation of data-driven models allowed a considerable reduction in sampling efforts while also providing process information for unsampled cultures. Therefore, the practical experiments of this study strongly affirm the previously claimed advantages of 2D fluorescence spectroscopy in microtiter plates.

rAAV Manufacturing: The Challenges of Soft Sensing during Upstream Processing

Recombinant adeno-associated virus (rAAV) is the most effective viral vector technology for directly translating the genomic revolution into medicinal therapies. However, the manufacturing of rAAV viral vectors remains challenging in the upstream processing with low rAAV yield in large-scale production and high cost, limiting the generalization of rAAV-based treatments. This situation can be improved by real-time monitoring of critical process parameters (CPP) that affect critical quality attributes (CQA). To achieve this aim, soft sensing combined with predictive modeling is an important strategy that can be used for optimizing the upstream process of rAAV production by monitoring critical process variables in real time. However, the development of soft sensors for rAAV production as a fast and low-cost monitoring approach is not an easy task. This review article describes four challenges and critically discusses the possible solutions that can enable the application of soft sensors for rAAV production monitoring. The challenges from a data scientist's perspective are (i) a predictor variable (soft-sensor inputs) set without AAV viral titer, (ii) multi-step forecasting, (iii) multiple process phases, and (iv) soft-sensor development composed of the mechanistic model.

Online 2D Fluorescence Monitoring in Microtiter Plates Allows Prediction of Cultivation Parameters and Considerable Reduction in Sampling Efforts for Parallel Cultivations of Hansenula polymorpha

Multi-wavelength (2D) fluorescence spectroscopy represents an important step towards exploiting the monitoring potential of microtiter plates (MTPs) during early-stage bioprocess development. In combination with multivariate data analysis (MVDA), important process information can be obtained, while repetitive, cost-intensive sample analytics can be reduced. This study provides a comprehensive experimental dataset of online and offline measurements for batch cultures of Hansenula polymorpha. In the first step, principal component analysis (PCA) was used to assess spectral data quality. Secondly, partial least-squares (PLS) regression models were generated, based on spectral data of two cultivation conditions and offline samples for glycerol, cell dry weight, and pH value. Thereby, the time-wise resolution increased 12-fold compared to the offline sampling interval of 6 h. The PLS models were validated using offline samples of a shorter sampling interval. Very good model transferability was shown during the PLS model application to the spectral data of cultures with six varying initial cultivation conditions. For all the predicted variables, a relative root-mean-square error (RMSE) below 6% was obtained. Based on the findings, the initial experimental strategy was re-evaluated and a more practical approach with minimised sampling effort and elevated experimental throughput was proposed. In conclusion, the study underlines the high potential of multi-wavelength (2D) fluorescence spectroscopy and provides an evaluation workflow for PLS modelling in microtiter plates.

Online optimization of dynamic binding capacity and productivity by model predictive control

In preparative and industrial chromatography, the current viewpoint is that the dynamic binding capacity governs the process economy, and increased dynamic binding capacity and column utilization are achieved at the expense of productivity. The dynamic binding capacity in chromatography increases with residence time until it reaches a plateau, whereas productivity has an optimum. Therefore, the loading step of a chromatographic process is a balancing act between productivity, column utilization, and buffer consumption. This work presents an online optimization approach for capture chromatography that employs a residence time gradient during the loading step to improve the traditional trade-off between productivity and resin utilization. The approach uses the extended Kalman filter as a soft sensor for product concentration in the system and a model predictive controller to accomplish online optimization using the pore diffusion model as a simple mechanistic model. When a soft sensor for the product is placed before and after the column, the model predictive controller can forecast the optimal condition to maximize productivity and resin utilization. The controller can also account for varying feed concentrations. This study examined the robustness as the feed concentration varied within a range of 50%. The online optimization was demonstrated with two model systems: purification of a monoclonal antibody by protein A affinity and lysozyme by cation-exchange chromatography. Using the presented optimization strategy with a controller saves up to 43% of the buffer and increases the productivity together with resin utilization in a similar range as a multi-column continuous counter-current loading process.

Convolutional Neural Network for Measurement of Suspended Solids and Turbidity

The great potential of the convolutional neural networks (CNNs) provides novel and alternative ways to monitor important parameters with high accuracy. In this study, we developed a soft sensor model for dynamic processes based on a CNN for the measurement of suspended solids and turbidity from a single image of the liquid sample to be measured by using a commercial smartphone camera (Android or IOS system) and light-emitting diode (LED) illumination. For this, an image dataset of liquid samples illuminated with white, red, green, and blue LED light was taken and used to train a CNN and fit a multiple linear regression (MLR) by using different color lighting, we evaluated which color gives more accurate information about the concentration of suspended particles in the sample. We implemented a pre-trained AlexNet model, and an MLR to estimate total suspended solids (TSS), and turbidity values in liquid samples based on suspended particles. The proposed technique obtained high goodness of fit (R2 = 0.99). The best performance was achieved using white light, with an accuracy of 98.24% and 97.20% for TSS and turbidity, respectively, with an operational range of 0–800 mgL−1, and 0–306 NTU. This system was designed for aquaculture environments and tested with both commercial fish feed and paprika. This motivates further research with different aquatic environments such as river water, domestic and industrial wastewater, and potable water, among others.

Oxygen Measurement in Microdevices

Oxygen plays a fundamental role in respiration and metabolism, and quantifying oxygen levels is essential in many environmental, industrial, and research settings. Microdevices facilitate the study of dynamic, oxygen-dependent effects in real time. This review is organized around the key needs for oxygen measurement in microdevices, including integrability into microfabricated systems; sensor dynamic range and sensitivity; spatially resolved measurements to map oxygen over two- or three-dimensional regions of interest; and compatibility with multimodal and multianalyte measurements. After a brief overview of biological readouts of oxygen, followed by oxygen sensor types that have been implemented in microscale devices and sensing mechanisms, this review presents select recent applications in organs-on-chip in vitro models and new sensor capabilities enabling oxygen microscopy, bioprocess manufacturing, and pharmaceutical industries. With the advancement of multiplexed, interconnected sensors and instruments and integration with industry workflows, intelligent microdevice-sensor systems including oxygen sensors will have further impact in environmental science, manufacturing, and medicine.

Ensemble-based adaptive soft sensor for fault-tolerant biomass monitoring

The accuracy and precision of soft sensors depend strongly on the reliability of underlying model inputs. These inputs (particularly readings of hardware sensors) are frequently subject to faults. This study aims to develop an adaptive soft sensor capable of reliable and robust biomass concentration predictions in the presence of faulty model inputs for a Pichia pastoris bioprocess. Hence, three soft sensor submodels were developed based on three independent model inputs (base addition, CO2 production, and mid-infrared spectrum). An ensemble-based algorithm combined the submodels to form an ensemble model, that is, an adaptive soft sensor, to achieve fault-tolerant prediction. The algorithm's basic steps are as follows: the initial determination of submodel reliability is followed by selecting appropriate submodels to generate a reliable prediction via variance-based weighting of the submodels. The adaptive soft sensor demonstrated high robustness and accuracy in biomass prediction in the presence of multiple simulated sensor faults (RMSE = 0.43 g L-1) and multiple real sensor faults (RMSE = 0.70 g L-1).

Modern Sensor Tools and Techniques for Monitoring, Controlling, and Improving Cell Culture Processes

The growing biopharmaceutical industry has reached a level of maturity that allows for the monitoring of numerous key variables for both process characterization and outcome predictions. Sensors were historically used in order to maintain an optimal environment within the reactor to optimize process performance. However, technological innovation has pushed towards on-line in situ continuous monitoring of quality attributes that could previously only be estimated off-line. These new sensing technologies when coupled with software models have shown promise for unique fingerprinting, smart process control, outcome improvement, and prediction. All this can be done without requiring invasive sampling or intervention on the system. In this paper, the state-of-the-art sensing technologies and their applications in the context of cell culture monitoring are reviewed with emphasis on the coming push towards industry 4.0 and smart manufacturing within the biopharmaceutical sector. Additionally, perspectives as to how this can be leveraged to improve both understanding and outcomes of cell culture processes are discussed.

Industrial data science – a review of machine learning applications for chemical and process industries

Understand and optimize industrial processes via machine learning and chemical engineering principles.