Model based strategies towards protein A resin lifetime optimization and supervision

Computer-aided design space identification for screening of protein A affinity chromatography resins

The rapidly growing market of monoclonal antibodies (mAbs) within the biopharmaceutical industry has incentivised numerous works on the design of more efficient production processes. Protein A affinity chromatography is regarded as one of the best processes for the capture of mAbs. Although the screening of Protein A resins has been previously examined, process flexibility has not been considered to date. Examining performance alongside flexibility is crucial for the design of processes that can handle disturbances arising from the feed stream. In this work, we present a model-based approach for the identification of design spaces, enhanced by machine learning. We demonstrate its capabilities on the design of a Protein A chromatography unit, screening five industrially relevant resins. The computational results favourably compare to experimental data and a resin performance comparison is presented. An improvement on the computational time by a factor of 300,000 is achieved using the machine learning aided methodology. This allowed for the identification of 5,120 different design spaces in only 19 h.

On-line PAT based monitoring and control of resin aging in protein A chromatography for COGs reduction

Resin aging is a common occurrence in chromatographic processes and generally influenced by factors such as cleaning procedure and composition of the feed stream. Two major events occur along with protein fouling, one is the loss of protein A ligand and the other is non-specific, irreversible interactions of foulants with resin particles. Both these are responsible for resin aging. As a result, the performance of the resin suffers a fall, and this can be quantified through indicators like reduction in dynamic binding capacity, increased column pressure, or peak broadening. The number of reuse cycles of a resin has a major influence on the cost per batch. This is even more significant in the case of protein A resin, which is the primary cost driver for downstream processing. In this work, we first identify chromatogram characteristics that correlate to resin aging. Next, we propose a data monitoring-based tool for prediction of resin aging. Principal component analysis of the UV data of Mab 1 showed a deviation at 120th cycle and an out of specification at around 149th cycle, corroborating with yield decline. Batch level modelling could deliver a predictable trend for resin aging and was demonstrated for two different Mabs (Mab1 and Mab2). The results demonstrate that significant resin aging can be detected 20-25 cycles prior to observable yield decline. A control strategy has been suggested such that once the deviation has been detected, additional resin cleaning is triggered. Overall, a 50-100 Protein A cycle enhancement in resin lifespan could be achieved.

Predictive mechanistic modeling of loading and elution in protein A chromatography

Protein A chromatography is an enabling technology in current manufacturing processes of monoclonal antibodies (mAbs) and mAb derivatives, largely due to its ability to reduce the levels of process-related impurities by several orders of magnitude. Despite its widespread application, the use of mathematical modeling capable of accurately predicting the full protein A chromatographic process, including loading, post-loading wash and elution stages, has been limited. This work describes a mechanistic modeling approach utilizing the general rate model (GRM), the capabilities of which are explored and optimized using two isotherm models. Isotherm parameters were estimated by inverse-fitting simulated breakthrough curves to experimental data at various pH values. The parameter values so obtained were interpolated across the relevant pH range using a best-fit curve, thus enabling their use in predictive modeling, including of elution over a range of pH. The model provides accurate predictions (< 3% mean error in 10% dynamic binding capacity predictions and ∼ 5% mean error in elution mass and pool volume predictions, both on scale-up) for various residence times, buffer conditions and elution schemes and its effectiveness for use in scale-up and process development is shown by applying the same parameters to larger columns and a wider range of residence times.

Exploring features in chromatographic profiles as a tool for monitoring column performance

In the biopharmaceutical industry, chromatography resins have a finite number of uses before they start to age and degrade, typically due to losses of ligand integrity and/or density. The "health" of a column is predicted and validated by running multiple cycles on representative scale-down models and can be followed by real-time on-going validation during commercial production. Principal Component Analysis (PCA), Partial Least Square (PLS), Similarity Scores and Single One Point-MultiParameter Technique (SOP-MPT) along with machine learning principles were applied to explore the hypothesis that there is predictive capability of latent variables in chromatography absorbance profiles for process performance (step yield) and product quality (aggregates, fragments, host cell proteins (HCP) and DNA, and Protein A ligand). The first stage of this study is described in this paper: a MabSelect SuRe™ chromatography column was cycled with a method to establish the "normal" baseline for process performance and product quality, followed by runs using a harsher NaOH Cleaning in Place (CIP) procedure (with a higher NaOH concentration than that recommended by the vendor) to accelerate resin degradation. The different mathematical analytical tools correlated with resin degradation of the column (reflected in decreasing step yield and binding capacity with increasing running cycle), specifically when using the Wash, Elution and Strip phases of the chromatography method. Monomer, HCP and DNA content were not significantly impacted and therefore a correlation with product quality was inconsequential. Importantly, this work shows proof-of-concept that while more traditional methods of measuring resin integrity such as the height equivalent to a theoretical place (HETP) and Asymmetry (As) measurements could not detect changes in the integrity of the resin, PCA, PLS, Similarity Scores and SOP-MPT (to a lesser extent) applied to the absorbance data were capable of anticipating issues in the chromatography bed by identifying atypical outcomes.

Process integrated biosensors for real-time monitoring of antibodies for automated affinity purification

Therapeutic monoclonal antibodies (mAbs) provide new means for treatments of a wide range of diseases and comprise a large fraction of all new approved drugs. Production of mAbs is expensive compared to conventional drug production, primarily due to the complex processes involved. The affinity purification step is dominating the cost of goods in mAb manufacturing. Process intensification and automation could reduce costs, but the lack of real-time process analytical technologies (PAT) complicates this development. We show a specific and robust fiber optical localized surface plasmon resonance (LSPR) sensor technology that is optimized for in-line product detection in the effluent in affinity capture steps. The sensor system comprises a flow cell and a replaceable sensor chip functionalized with biorecognition elements for specific analyte detection. The high selectivity of the sensor enable detection of mAbs in complex sample matrices at concentrations below 2.5 μg mL^-1. In place regeneration of the sensor chips allowed for continuous monitoring of multiple consecutive chromatographic separation cycles. Excellent performance was obtained at different purification scales with flow rates up to 200 mL min^-1. This sensor technology facilitates efficient column loading, optimization, and control of chromatography systems, which can pave the way for continuous operation and automation of protein purification steps.

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.

Scalable, Robust and Highly Productive Novel Convecdiff Membrane Platform for mAb Capture

The recombinant monoclonal antibody capture step represents the current bottleneck in downstream processing. Protein A resins are diffusion-limited chromatography materials which require low flow rates to achieve a binding capacity above 30 g L⁻¹ with the result of low productivity. Here, we present a novel chromatography membrane combining superior binding capacities with high flow rates for high productivity while achieving comparable product quality as state-of-the-art protein A resins. Further, we demonstrate full scalability of this convecdiff technology with experimental data demonstrating suitability for bioprocessing at different scales. This technology results in more than 10-fold higher productivity compared to Protein A resins, which is maintained during scale up. We demonstrate the influence of residence times, feed titers and the cleaning regime on productivity and indicate optimal utilization of the convecdiff membrane based on feed titer availability. The underlying high productivity and short cycle times of this material enable the purification of monoclonal antibodies with 10-times less chromatography material used per batch and utilization of the membrane within one batch. Provided in disposable consumables, this novel technology will remove column handling in bioprocesses and resin re-use over multiple batches.

A regressive approach to the design of continuous capture process with multi-column chromatography for monoclonal antibodies

Although empirical methods have been introduced in the process development of continuous chromatography, the common approach to optimize a multi-column continuous capture chromatography (periodic counter-current chromatography, PCCC) process heavily relies on numerical model simulations and the number of experiments. In addition, different multi-column settings in PCCC add more design variables in process development. In this study, we have developed a rational method for designing PCCC processes based on iterative calculations by mechanistic model-based simulations. Breakthrough curves of a monoclonal antibody were measured at different residence times for three protein A resins of different particle sizes and capacities to obtain the parameters needed for the simulation. Numerical calculations were performed for the protein sample concentration in the range of 1.5 to 4 g/L. Regression curves were developed to describe the relative process performances compared with batch operation, including the resin capacity utilization and the buffer consumption. Another linear correlation was established between breakthrough cut-off (BT%) and a modified group composed of residence time, mass transfer coefficient, and particle size. By normalizing BT% with binding capacity and switching time, the linear regression curves were established for the three protein A resins, which are useful for the design and optimization of PCCC to reduce the process development time.

Hybrid Models for the simulation and prediction of chromatographic processes for protein capture

The biopharmaceutical industries are continuously faced with the pressure to reduce the development costs and accelerate development time scales. The traditional approach of heuristic-based or platform process-based optimization is soon getting obsolete, and more generalized tools for process development and optimization are required to keep pace with the emerging trends. Thus, advanced model-based methods that can reduce the can ensure accelerated development of robust processes with minimal experiments are necessary. Though mechanistic models for chromatography are quite popular, their success is limited by the need to have accurate knowledge of adsorption isotherms and mass transfer kinetics. As an alternative, in this work, a hybrid modeling approach is proposed. Thereby, the chromatographic unit behavior is learned by a combination of neural network and mechanistic model while fitting suitable experimental breakthrough curves. Since this approach does not require identifying suitable mechanistic assumptions for all the phenomena, it can be developed with lower effort. Thus, allowing the scientists to concentrate their focus on process development. The performance of the hybrid model is compared with the mechanistic Lumped kinetic Model for in-silico data and experiments conducted on a system of industrial relevance. The flexibility of the hybrid modeling approach results in about three times higher accuracies compared to Lumped Kinetic Model. This is validated for five different isotherm models used to simulate data, with the hybrid model showing about two to three times lower prediction errors in all the cases. Not only in prediction, but we could also show that the hybrid model is more robust in extrapolating across process conditions with about three times lower error than the LKM. Additionally, it could be demonstrated that an appropriately tailored formulation of the hybrid model can be used to generate representations for the underlying principles such as adsorption equilibria and mass transfer kinetics.