Downstream
Presentations in the “Recovery and Purification” track of the 2024 BPI Conference and Exhibition emphasized the need for increased understanding of downstream processing (DSP) for both conventional and novel therapeutics. Historically, cultivating sufficient knowledge — e.g., to screen candidate adsorbents for affinity capture of a nonstandard protein — would have involved considerable time, resources, and experimental efforts. But downstream scientists are beginning to develop critical insights by leveraging recent advances in computational modeling and data analytics. Presentations in the DSP track demonstrated the promise of in silico tools for enhancing the effectiveness and process economics of purification and filtration steps.
Some presenters highlighted applications of machine learning (ML) to enhance chromatography-resin screening, particularly for resins with mixed-mode chemistries. Steven Cramer (a professor of chemical and biological engineering at Rensselaer Polytechnic Institute) reported on his team’s efforts to identify molecular descriptors for a set of therapeutic proteins. Such descriptors, he explained, will be essential for training ML-based tools to predict resin selectivity based on product-protein characteristics — e.g., hydrophobicity patches and preferred binding regions. Cramer’s group seeks to combine insights obtained from wet-laboratory studies of separation behavior and biophysical simulations, ultimately with the goal of facilitating adsorbent evaluation.
Nick Vecchiarello (assistant professor of chemical engineering at the University of Virginia) focused on strategies for modeling and measuring partition coefficient (Kp) values, which DSP scientists use during early and late-stage process development to evaluate the binding, elution, and selectivity behaviors of candidate resins and operating conditions. Specifically, Vecchiarello provided a framework to inform the selection of phase-ratio and protein-loading parameters. He also introduced a software tool that his team developed to calculate operating regimes. Freely available through the GitHub platform, the program enables researchers to use in-house isothermic data to calculate custom operating regimes.
Progress toward mechanistic modeling of DSP steps also was featured prominently in the track. Lei Wang (senior scientist II at AbbVie) spoke about application of GoSilico modeling tools (Cytiva) during early purification-process development. In one study, his team compared results from GoSilico modeling of a cation-exchange (CEX) polishing step for a conventional monoclonal antibody (mAb) with data from wet-laboratory experiments, ultimately demonstrating the model’s accuracy and feasibility. The team also modeled clearance of particular host-cell impurities from a bispecific antibody (bsAb) product. Wang observed that, despite some modeling challenges attributable to protein–ligand secondary interactions, such tools represent an important step forward in applying first principles to facilitate early process-development activities.
Harin Ozbakir (a senior scientist in process development at Amgen) called attention to the need for mechanistic modeling in the context of reagent clearance. He focused on assessment of clearance during ultrafiltration/diafiltration (UF/DF). Although such processes are well understood, working toward mechanistic understanding of reagent clearance during UF/DF would help to optimize the time and experimental effort required to identify optimal process conditions. The Amgen team spiked a drug substance lacking in charged species with known concentrations of two reagents, then performed a two-stage UF/DF process to measure clearance. The team also applied two in silico approaches based on first principles for mass transfer and protein-charge calculations, respectfully. Ozbakir observed that leveraging confirmatory modeling during early experiments could reduce required time and experimental effort by up to 80% for such testing. As such presentations suggest, bringing data-driven approaches to bear on DSP could help to improve screening and optimize the resources required to demonstrate the effectiveness of critical downstream steps.
The diversity of monoclonal antibody therapeutics has expanded significantly in recent years resulting in new challenges for downstream processing. Here, we present 2 studies to highlight our toolbox approach to significantly reduce impurities such as host cell proteins, aggregates, and endotoxin, and achieve viral clearance from widely used expression systems.
Study 1: We demonstrate significant reduction of impurities, while achieving high target protein recovery by using a novel mixed mode chromatography resin, HCPure™, designed to operate in a unique chemical space.
Study 2: We examine the potential of 2 orthogonal adsorbents, HCPure™ and Q PuraBead® to remove noninfectious viral surrogates for Minute Virus of Mice and retrovirus (Cygnus Technologies).
Caroline Daye is a Research Scientist at Astrea Bioseparations, a cutting-edge biotechnology company specializing in advanced chromatography solutions. Following her BSc in Biology at Sheffield Hallam University, Caroline started her professional journey as a Research Assistant at The University of Sheffield Medical school Oncology department before joining BioServUK specialising in the production and purification of IgG from hybridoma cell lines. Caroline has been with Astrea Bioseparations for the last 6 years, where she plays an important role in both downstream bioprocessing and analysis.
