The emergence of diverse biotherapeutic modalities, ranging from monoclonal antibodies (mAbs) to multispecific biologics and antibody-drug conjugates (ADCs), has revolutionized drug discovery. In this presentation, I will demonstrate how we integrate high-throughput techniques and automated processes to accelerate the characterization of novel biotherapeutics. By leveraging advanced analytical platforms, automation technologies, and data-driven approaches, we can efficiently assess quality attributes, deepen product understanding, and streamline the development of transformative new medicines.

Successful antibody therapeutics are dependent on a myriad of properties including potency, specificity, and pharmacokinetics. This work highlights their interplay via an affinity optimization campaign targeting species cross-reactivity. Antibodies with improved cross-reactivity showed high levels of off-target polyreactivity. These trade-offs were overcome in a subsequent yeast display campaign, which provided mutations for pharmacokinetic engineering. Integrating these mutations into the variable region, we were able to modulate pharmacokinetic behavior.

The demand for high-throughput intact analysis has steadily increased with implementation of high-throughput expression and purification. Analyzing proteins at the intact level offers a fast and straightforward characterization approach for QC. Implementing ProteinMetrics workflow can accelerate data processing to around 1 minute/sample, resulting in a throughput of more than 3000 samples/week. The software takes protein sequences as input, calculates theoretical molecular weight, and automatically annotates peaks in the spectrum.

Subcutaneous administration of biologics necessitates high-concentration formulations, often leading to increased viscosity, complicating development and administration. This study introduces a multimodal feature-based machine learning workflow for predicting antibody viscosity. By integrating diverse data types—sequence, structural, physiochemical, and language model embeddings—the workflow learns from physiochemical and protein evolutionary rules. This approach aims to streamline high-concentration biologic development, enhancing therapeutic efficacy and patient compliance.

Biotherapeutic engineering campaigns require high-throughput assays to design proteins with favorable developability and functional properties. Implementing these assays in an early research context requires a harmonization of production, purification, and analytical techniques to rapidly generate the proper data to drive project decisions. I will describe our strategy for implementing high-throughput analytics in our engineering workflows with focus on sample consumption, throughput, and maximization of analytical diversity. 

Biotherapeutic manufacturing processes subject proteins to a variety of potentially destabilizing conditions, including but not limited to shifts in temperature, pH, and chemical environment, and interfacial stress. We present a manufacturability suite of scale-down methods that can be used to identify liabilities and prioritize candidates when transitioning from research to development. Our development and benchmarking of these methods reveals liabilities unique to each IgG subclass and light chain isotype.

Two case studies covering biophysical characterization of protein solutions will be presented. First, water proton nuclear magnetic resonance (wNMR) was employed as a noninvasive, in situ method to assess aggregation propensity of a high-concentration drug product, Dupixent®, directly in primary containers. Second, a Reciprocal Injection Device (RID) was introduced to accelerate screening of concentration-dependent protein stability and interfacial interactions of formulations under intensified stress conditions within prefilled syringes.

• Realistic Expectations: Where does AI/ML provide the most value and where does it fall short in bioprocessing
• Feasibility: How practical are these tools for biologics manufacturing?
• Upstream Applications: Can AI/ML improve cell culture optimization, media selection, and yield prediction?
• Downstream Relevance: How AI can assist with purification, process scaling, and troubleshooting bottlenecks?
• Analytics and Monitoring: Using AI to predict product stability, aggregation, and monitor CQAs in continuous manufacturing.
• Integration Strategies: Practical tips for validating and embedding AI/ML models into existing workflows.​

• High throughput analytics for developability workflow
• Immunogenicity prediction and ex vivo assays to assess immunogenicity
• Aggregation prediction and analytical technologies to assess aggregation propensity 
• PTM prediction and analytical tools to assess molecular liability 
• Balance between biological function and biophysical stability​

The presented antibody developability assessment approach aims to identify and characterize lead candidates which exhibit favorable biophysical properties, thereby reducing risk in the drug development process and identifying pathways for lead candidates to become stable drug products. Using commercially available mAbs, we evaluate the utility of novel biophysical methods that complement established approaches. We demonstrate that unfolding reversibility by Modulated Scanning Fluorimetry (MSF), chemical denaturation and refolding, and the concentration dependency of Tm (∆Tm) provide valuable insights when selecting the most promising candidates for mAb development.

Nano differential scanning fluorimetry (nanoDSF) is a high-throughput screening technique for simultaneous evaluation of colloidal and conformational stability. However, this technique has not been explored enough for studying the aggregation behavior of the formulations. We evaluated three different vaccine formulations and their structural integrity and colloidal behavior using defined physiological conditions and have shown the potential of nano-DSF technology for process optimization, to gain the product knowledge, process improvement and to use as a preliminary decision-making tool before a long-term stability study is initiated.

Process design is challenging when one operation's outcome significantly impacts the next. In precipitation-based processes, the 3D structure of precipitates plays a crucial role in recovery and dissolution. This presentation introduces new high-throughput methods to screen and characterize precipitates and precipitation conditions. Furthermore, we will also present our integrated and continuous manufacturing process for mRNA production and purification. Additionally, prediction models for process development will be discussed.

The emergence of highly potent therapeutics with low expected clinical doses creates a challenge for analytical characterization of simulated drug product in-use samples. Sample characterization for protein concentration, size variants, charge variants, and potency often necessitates additional analytical method development to improve sensitivity and ensure compatibility with in-use samples. Here we report the development and qualification of reliable in-use methods to characterize simulated in-use samples during drug product development.

Single-cell characterization is one of the most critical measurements—being utilized in medical diagnostics, cellular therapeutics, and biomanufacturing applications. Different cellular-therapies-based biomanufacturing assays require single-cell monitoring while determining cellular potency, viability, and activity. Single-cell measurements serve as QC in many biomanufacturing processes. Here, I will present our recently 3D-printed, autonomous bioanalytical imaging and characterization setup based on florescence microscopy, capable of imaging cells at point-of-care.