Digital twin of a continuous chromatography process for mAb purification: Design and model-based control.

Model-based design of integrated continuous train coupled with online process analytical technology (PAT) tool can be a potent facilitator for monitoring and control of Critical Quality Attributes (CQAs) in real time. Charge variants are product related variants and are often regarded as CQAs as they may impact potency and efficacy of drug. Robust pooling decision is required for achieving uniform charge variant composition for mAbs as baseline separation between closely related variants is rarely achieved in process scale chromatography. In this study, we propose a digital twin of a continuous chromatography process, integrated with an online HPLC-PAT tool for delivering real time pooling decisions to achieve uniform charge variant composition. The integrated downstream process comprised continuous multicolumn capture protein A chromatography, viral inactivation in coiled flow inverter reactor (CFIR), and multicolumn CEX polishing step. An online HPLC was connected to the harvest tank before protein A chromatography. Both empirical and mechanistic modeling have been considered. The model states were updated in real time using online HPLC charge variant data for prediction of the initial and final cut point for CEX eluate, according to which the process chromatography was directed to switch from collection to waste to achieve the desired charge variant composition in the CEX pool. Two case studies were carried out to demonstrate this control strategy. In the first case study, the continuous train was run for initially 14 h for harvest of fixed charge variant composition as feed. In the second case study, charge variant composition was dynamically changed by introducing forced perturbation to mimic the deviations that may be encountered during perfusion cell culture. The control strategy was successfully implemented for more than ±5% variability in the acidic variants of the feed with its composition in the range of acidic (13%-17%), main (18%-23%), and basic (59%-68%) variants. Both the case studies yielded CEX pool of uniform distribution of acidic, main and basic profiles in the range of 15 ± 0.8, 31 ± 0.3, and 53 ± 0.5%, respectively, in the case of empirical modeling and 15 ± 0.5, 31 ± 0.3, and 53 ± 0.3%, respectively, in the case of mechanistic modeling. In both cases, process yield for main species was >85% and the use of online HPLC early in the purification train helped in making quicker decision for pooling of CEX eluate. The results thus successfully demonstrate the technical feasibility of creating digital twins of bioprocess operations and their utility for process control.

Control of surge tanks for continuous manufacturing of monoclonal antibodies.

Surge tanks are critical but often overlooked enablers of continuous bioprocessing. They provide multiple benefits including dampening of concentration gradients and allowing process resumption efforts in case of equipment failure or unexpected deviations, which can occur during a continuous campaign of weeks or months. They are also useful in enabling steady-state operation across a continuous train by facilitating mass balance between unit operations such as chromatography which have periodic loading and elution cycles. In this paper, we propose a design of a system of surge tanks for a monoclonal antibody (mAb) production process consisting of cell culture, clarification, capture chromatography, viral inactivation, polishing chromatography, and single-pass ultrafiltration and diafiltration. A Python controller has been developed for robust control of the continuous train. The controller has four layers, namely data acquisition, process scheduling, deviation handling, and real-time execution. A set of general guidelines for surge tank placement and sizing have been proposed together with process control strategies based on the design space of the individual unit operations, failure modes analysis of the different equipment, and expected variability in the process feed streams for both fed-batch and perfusion bioreactors. The control system has been successfully demonstrated for several continuous runs of up to 36 h in duration and is able to leverage surge tanks for robust control of the continuous train in the face of product variability as well as process errors while maintaining critical quality attributes. The proposed set of strategies for surge tank control are adaptable to most continuous processing setups for mAbs, and together form the first framework that can fully realize the benefits of surge tanks in continuous bioprocessing.

Use of HPLC as an Enabler of Process Analytical Technology in Process Chromatography.

Online monitoring of product quality attributes using high resolution analytical tools is a prerequisite for implementation of process analytical technology (PAT) and thereby ensuring product quality and consistency. Online high-performance liquid chromatography (HPLC) has been established for real time monitoring of product quality attributes. However, requirement of liquid handling system capable of online sampling and fractionation and interfacing it with preparative scale chromatography appends to the cost and complexity of the design module of commercially available online HPLC. This paper proposes a cost-effective approach for using a traditional offline HPLC for online analysis using a 2 way/6 port valve to facilitate simultaneous automated sampling of product stream eluting from a process column and fractionation. No sample dilution is required in the proposed approach. The versatility of the proposed online configuration has been verified by demonstrating its use for two of the most common separations required during production of monoclonal antibody therapeutics: separation of aggregates and separation of charge variants. Process modeling has been performed to allow interpolation of HPLC data and facilitate pooling to achieve the desired purity (model predicted purity 99.1% vs achieved purity of 99.3% for removal of aggregates, model predicted main species yield of 62.4% vs achieved main species of 62.9% for pooling of charge variants). The study thus demonstrates that the proposed online HPLC configuration can be used for PAT applications in preparative chromatography.

Application of advanced machine learning algorithms for anomaly detection and quantitative prediction in protein A chromatography.

Protein A capture chromatography, which forms the core of the mAb purification platform, demands cautious use and maximum resin utilization due to high cost associated with resin. In this paper, we propose an application of advanced machine learning (ML) algorithms to address two most crucial objectives of column integrity breach and yield prediction for resin cycling study of protein A chromatography. Two approaches have been considered to detect anomalies in case of column integrity breach. The first approach utilized the traditional Principal Component Analysis (PCA) method for dimensionality reduction followed by anomaly detection using Isolation Forest (IF) algorithm. The second approach involved the application of deep learning neural network based Long Short Term Memory autoencoder (LSTM AE). Both the algorithms could successfully predict the column integrity failure 4 cycles ahead of the actual cycle. In the case of prediction of percentage yield decay, a partial least squares-artificial neural network (PLS-ANN) augmented model was utilized and compared with the traditional PLS regression model. The developed PLS-ANN model with higher R2 and lower RMSE values of 0.96 and 0.014 respectively could outperform the classical PLS model with lower R2 and RMSE values of 0.88 and 0.028, resulting in more accurate yield prediction. The developed ML algorithms for both case studies could not only successfully forecast anomalies by detecting subtle changes in column packing quality and thereby facilitate real time control decisions for preventive measures, a prerequisite for continuous manufacturing, but also demonstrated the ability to predict complex yield decay behaviour for protein A chromatography. As biopharmaceutical manufacturing adopts continuous processing, copious amount of data will be generated from the process and analytical equipment on the manufacturing floor, and the proposed advanced ML algorithms have significant potential in dealing with nonlinearities of the different unit operations simultaneously and facilitate real-time control decision making.

An accelerated approach for mechanistic model based prediction of linear gradient elution ion-exchange chromatography of proteins.

With growing demands for therapeutic monoclonal antibodies, in silico downstream process development based on mechanistic modeling of chromatography separation process is being increasingly used for process optimization and process characterization. Application of mechanistic modeling in biopharmaceutical industry has been sparse due to the significant investment of time and resources that are required for performing model calibration. Mechanistic modeling of the chromatography process involves a large number of mass transport and binding parameters and their initial input values are required for simulations. These input values of column parameters can be easily obtained either from experiments or from empirical correlations available in literature. On the other hand, obtaining the model input valves for binding kinetic parameters is usually a cumbersome process as it involves performing batch experiments which are not only tedious but also require significant quantities of purely isolated main product and its related impurities, which is challenging as the product related impurities are typically present in smaller quantities and hence are difficult to obtain as pure species. In the present work, a mechanistic model that is based on the general rate model coupled with extended Langmuir binding model has been used for prediction of linear gradient elution peaks of monoclonal antibody on cation exchanger chromatography. The present work describes an accelerated approach for obtaining the input values for binding kinetic parameters in the extended Langmuir binding model from the two Yamamoto coefficient A and B values obtained by Yamamoto method directly from the model calibration linear gradient elution runs of different gradient slopes and at low to moderate protein loadings. The equations that can relate the two coefficients to the extended Langmuir model equation binding kinetic parameters were derived. Therefore, once A and B are determined, the binding kinetic parameter values were determined straightforward, thereby avoiding the problem of multiple solutions for the model parameters. The estimated binding parameters were successfully validated from isocratic elution experiments performed at low loading. What we demonstrate is that the proposed approach allows us to estimate binding kinetic parameters in a significantly more efficient and accelerated manner than presently used approaches, thereby accelerating development and implementation of mechanistic modeling for process chromatography.

Complete or periodic continuity in continuous manufacturing platforms for production of monoclonal antibodies?

BACKGROUND: Monoclonal antibodies (mAbs) currently dominate the biotherapeutic market. This has resulted in significant efforts towards the development of a continuous integrated platform for the manufacturing of mAbs. MAIN METHODS AND MAJOR RESULTS: In this study, a continuous mAb platform has been developed consisting of an Acoustic Wave Separator, a Cadence BioSMB PD system, a customized coiled flow reactor, a modular single-pass TFF kit, an in-line diafiltration module, and a continuous dead-end filtration skid. A three-step chromatographic purification was performed in the platform consisting of Protein A capture chromatography followed by an anion exchange membrane directly coupled to a cation exchange chromatography. Two operational case studies have been executed on the platform, namely complete continuous ("CC") and periodic continuous ("PC") modes of operation. The CC mode was designed to ensure that each unit operation had completely continuous inflow and outflow by increasing the number of columns, filtration modules and tanks, while the PC mode operated in periodic pulses with scheduled flow and hold steps. Both modes were designed to handle the same flow rate and titers from the upstream bioreactor or fed-batch harvest tank, and were compared in terms of productivity and operational complexity. Both modes offer viable options for continuous processing of mAbs and result in achievement of target critical quality attribute profiles of the final drug product over 24 h of operation. CONCLUSIONS AND IMPLICATIONS: It was found that the CC mode was superior in terms of specific productivity (20-50% higher) and consumable utilization (20% lower resin utilization), while the PC mode was operationally simpler and had lower facility costs due to significant reductions in the number of auxiliary equipment (pumps, columns, tanks, and valves). The work successfully highlighted the pros and cons of both approaches, and demonstrates that while several groups have amply shown the superiority of continuous processing over batch mode, there are intermediate variants which may be optimal in a given situation.

Retinopathy as a Prognostic Marker in Cerebral Malaria.

OBJECTIVE: To study the association between fundal changes (malarial retinopathy) and mortality in children with cerebral malaria. METHODS: 50 consecutive children (mean age 8.4 y, 23 males) with cerebral malaria (acute febrile encephalopathy and either peripheral smear or Rapid diagnostic test positive for malaria) were evaluated by a single ophthalmologist for any changes of retinopathy. Children were managed as per standard guidelines for treatment of cerebral malaria. RESULTS: P. vivax infection was seen in one child, P. falciparum infection in 42 children, and a mixed infection in 7. Retinopathy was present in 48% of the children. 13 children died during hospital stay. The mean interval from admission to fundus examination was 11.6 (4.64) h. Presence of any retinopathy (P=0.02), and either of papilledema (P=0.02), hemorrhages (P=0.005) or vessel changes (P=0.01), were associated with a significantly higher risk of death. CONCLUSION: Malarial retinopathy is significantly associated with mortality in children with cerebral malaria. It may be used for both prognostication, and triaging for optimum utilization of intensive care facilities in these children.

Use of ultrasound biomicroscopy to predict long-term outcome of sub-Tenon needle revision of failed trabeculectomy blebs: a pilot study.

PURPOSE: To evaluate the role of ultrasound biomicroscopy (UBM) in predicting the long-term outcome of sub-Tenon needling revision of failed trabeculectomy blebs. METHODS: Adult patients with a failed trabeculectomy bleb and unsatisfactory intraocular pressure (IOP) control were recruited. The aqueous flow under the partial thickness scleral flap was looked for and the blebs classified on UBM as scleral route patent (SRP) or scleral route occluded (SRO). All blebs underwent needling revision with injection of 5 mg/0.1 mL 5-fluorouracil. Survival of the revision procedure at the end of 2 years follow-up with regards to the baseline UBM characterization of the bleb was noted. Successful outcome was defined as IOP <22.0 mmHg and/or 30% reduction of baseline IOP with or without medication. RESULTS: A total of 13 eyes had SRP and 5 eyes had SRO blebs on UBM. Only SRP blebs survived the needling procedure by the end of 2 years. Of the 13 SRP blebs, 10 blebs survived (76.9%). Needling had failed in all 5 SRO blebs. The overall success rate was 55.6% at 2 years. There was no difference in age, IOP, and time from initial trabeculectomy between the failed and successful group. The outcome correlated significantly to the patency of the scleral route assessed by UBM (p=0.07). CONCLUSIONS: Ultrasound biomicroscopy characterization of failed blebs appears to help in predicting the outcome of needle revision. In SRO blebs, it may be better to plan a full bleb revision rather than needling alone. Ultrasound biomicroscopy may help in avoiding an unnecessary needling procedure in SRO blebs where it is likely to fail.