The protein cloud-point temperature (TCloud) is a known representative of protein–protein interaction strength and provides
valuable information during the development and characterization of protein-based products, such as biopharmaceutics. A
high-throughput low volume TCloud detection method was introduced in preceding work, where it was concluded that the
extracted value is an apparent TCloud (TCloud,app). As an understanding of the apparent nature is imperative to facilitate inter-
study data comparability, the current work was performed to systematically evaluate the influence of 3 image analysis strate-
gies and 2 experimental parameters (sample volume and cooling rate) on TCloud,app detection of lysozyme. Different image
analysis strategies showed that TCloud,app is detectable by means of total pixel intensity difference and the total number of
white pixels, but the latter is also able to extract the ice nucleation temperature. Experimental parameter variation showed a
TCloud,app depression for increasing cooling rates (0.1–0.5 °C/min), and larger sample volumes (5–24 μL). Exploratory ther-
mographic data indicated this resulted from a temperature discrepancy between the measured temperature by the cryogenic
device and the actual sample temperature. Literature validation confirmed that the discrepancy does not affect the relative
inter-study comparability of the samples, regardless of the image analysis strategy or experimental parameters. Additionally,
high measurement precision was demonstrated, as TCloud,app changes were detectable down to a sample volume of only 5 μL
and for 0.1 °C/min cooling rate increments. This work explains the apparent nature of the TCloud detection method, showcases
its detection precision, and broadens the applicability of the experimental setup

Image-based protein phase diagram analysis is key for understanding and exploiting protein phase behavior in the biopharmaceutical field. However, required data analysis has become a notorious time-consuming task since high-throughput screening approaches were implemented. A variety of computational tools have been developed to support analysis, but these tools primarily use end point visible light images. This study investigates the combined effect of end point and time-dependent image features obtained from cross-polarized and ultraviolet light features, supplementary to visible light, on protein phase diagram image classification. In addition, external validation was performed to evaluate the classification algorithm's applicability to support protein phase diagram scoring. The predicted protein phase behavior classes were subsequently used to automatically construct multidimensional protein phase diagrams to prevent image information loss without complicating the used image classification algorithm. Combining end point and time-dependent features from 3 light sources resulted in a balanced accuracy of 86.4 ± 4.3%, which is comparable to or better than more complex classifiers reported in literature. External validation resulted in a correct formulation classification rate of 91.7%. Subsequent automated construction of the multidimensional protein phase diagrams, using predicted classes, allowed visualization of details such as crystallization rate and protein phase behavior type coexistence.

Abstract: Short-term parameters correlating to long-term protein stability, such as the protein cloud point temperature (T_cloud), are of interest to improve efficiency during protein product development. Such efficiency is reached if short-term parameters are obtained in a low volume and high-throughput (HT) manner. This study presents a low volume HT detection method for (sub-zero) T_cloud determination of lysozyme, as such an experimental method is not available yet. The setup consists of a cryogenic device with an automated imaging system. Measurement reproducibility (median absolute deviation of 0.2 °C) and literature-based parameter validation (Pearson correlation coefficient of 0.996) were shown by a robustness and validation study. The subsequent case study demonstrated a partial correlation between the obtained apparent T_cloud parameter and long-term protein stability as a function of lysozyme concentration, ion type, ionic strength, and freeze/thaw stress. The presented experimental setup demonstrates its ability to advance short-term strategies for efficient protein formulation development. Graphical Abstract: [Figure not available: see fulltext.].

Redesigning existing food protein formulations is necessary in situations where food authorities propose dose adjustments or removal of currently employed additives. Redesigning formulations involves evaluating substitute additives to obtain similar long-term physical stability as the original formulation. Such formulation screening experiments benefit from comprehensive data visualization, understanding the effects of substitute additives on long-term physical stability, and identification of short-term optimization targets. This work employs empirical phase diagrams to reach these benefits by combining multidimensional long-term protein physical stability data with short-term empirical protein properties. A case study was performed where multidimensional protein phase diagrams (1152 formulations) allowed for identification of stabilizing effects as a result of pH, methionine, sugars, salt, and minimized glycerol content. Corresponding empirical protein property diagrams (144 formulations) resulted in the identification of normalized surface tension as a short-term empirical protein property to reach long-term physical stability presumably similar to the original product, namely via preferential hydration. Additionally, changes in pH and salt were identified as environmental optimization targets to reach stability via repulsive electrostatic forces. This case study shows the applicability of the empirical phase diagram method to rationally perform formulation redesign screenings, while simultaneously expanding knowledge on protein long-term physical stability.

Knowledge-based experimental design can aid biopharmaceutical high-throughput screening (HTS) experiments needed to identify critical manufacturability parameters. Prior knowledge can be obtained via computational methods such as protein property extraction from 3-D protein structures. This study presents a high-throughput 3-D structure preparation and refinement pipeline that supports structure screenings with an automated and data-dependent workflow. As a case study, three chimeric virus-like particle (VLP) building blocks, hepatitis B core antigen (HBcAg) dimers, were constructed. Molecular dynamics (MD) refinement quality, speed, stability, and correlation to zeta potential data was evaluated using different MD simulation settings. Settings included 2 force fields (YASARA2 and AMBER03) and 2 pKa computation methods (YASARA and H++). MD simulations contained a data-dependent termination via identification of a 2 ns Window of Stability, which was also used for robust descriptor extraction. MD simulation with YASARA2, independent of pKa computation method, was found to be most stable and computationally efficient. These settings resulted in a fast refinement (6.6–37.5 h), a good structure quality (−1.17–−1.13) and a strong linear dependence between dimer surface charge and complete chimeric HBcAg VLP zeta potential. These results indicate the computational pipeline's applicability for early-stage candidate assessment and design optimization of HTS manufacturability or formulability experiments.

Identification of long-term stable biopharmaceutical formulations is essential for biopharmaceutical product development. Reduction of the number of long-term storage experiments and a well-defined formulation search space requires knowledge-based formulation screenings and a detailed protein phase behavior understanding. To achieve this, short-term analytical techniques can serve as predictors for long-term protein phase behavior. Protein phase behavior studies that investigate this concept commonly display shortcomings such as limited and small datasets, sample adjustments, or simplistic data analysis. To overcome these shortcomings, 150 unique lysozyme solutions were analyzed using six different short-term analytical techniques. Lysozyme's structural properties, conformational stability, colloidal stability, surface charge, and surface hydrophobicity were obtained directly after formulation preparation. Employing the empirical phase diagram method, this short-term data was correlated to long-term physical stability data obtained during 40 days of storage. Short-term protein properties showed partial correlation to long-term phase behavior. Structural differences, changing surface properties, colloidal stability, and conformation stability as a function of formulation conditions were observed. This study contributes to long-term protein phase behavior research by presenting a systematic, data-dependent, and multidimensional data evaluation workflow to create a comprehensive overview of short-term protein analytics in relation to long-term protein phase behavior.

Protein phase diagrams are a tool to investigate the cause and consequence of solution conditions on protein phase behavior. The effects are scored according to aggregation morphologies such as crystals or amorphous precipitates. Solution conditions affect morphologic features, such as crystal size, as well as kinetic features, such as crystal growth time. Commonly used data visualization techniques include individual line graphs or phase diagrams based on symbols. These techniques have limitations in terms of handling large data sets, comprehensiveness or completeness. To eliminate these limitations, morphologic and kinetic features obtained from crystallization images generated with high throughput microbatch experiments have been visualized with radar charts in combination with the empirical phase diagram method. Morphologic features (crystal size, shape, and number, as well as precipitate size) and kinetic features (crystal and precipitate onset and growth time) are extracted for 768 solutions with varying chicken egg white lysozyme concentration, salt type, ionic strength, and pH. Image-based aggregation morphology and kinetic features were compiled into a single and easily interpretable figure, thereby showing that the empirical phase diagram method can support high-throughput crystallization experiments in its data amount as well as its data complexity.

Chromatographic separation of biopharmaceuticals in general and monoclonal antibodies (mAbs) specifically is the bottleneck in terms of cost and throughput in preparative purification. Still, generalized platform processes are used, neglecting molecule specific characteristics, defining protein-resin interaction terms. Currently used in silico modeling approaches do not consider the orientation of the molecule towards the chromatographic resins as a result of the structural features on an atomic level. This paper describes a quantitative structure–activity relationship (QSAR) approach to model the orientation of mAbs on ion exchange chromatographic matrices as a function of property distribution and mobile phase characteristics. 6 mAbs were used to build a predictive QSAR model and to investigate the preferred binding orientations and resulting surface shielding on resins. Thereby different dominating orientations, caused by composition of F_ab fragments of the mAbs, could be identified. The presented methodology is suitable to gain extended insight in molecule orientation on chromatographic resins and to tailor purification strategies based on molecule structure.

Quantitative structure–activity relationship (QSAR) modeling for prediction of biomolecule parameters has become an established technique in chromatographic purification process design. Unfortunately available descriptor sets fail to describe the orientation of biomolecules and the effects of ionic strength in the mobile phase on the interaction with the stationary phase. The literature describes several special descriptors used for chromatographic retention modeling, all of these do not describe the screening of electrostatic potential by the mobile phase in use. In this work we introduce two new approaches of descriptor calculations, namely surface patches and plane projection, which capture an oriented binding to charged surfaces and steric hindrance of the interaction with chromatographic ligands with regard to electrostatic potential screening by mobile phase ions. We present the use of the developed descriptor sets for predictive modeling of Langmuir isotherms for proteins at different pH values between pH 5 and 10 and varying ionic strength in the range of 10–100 mM. The resulting model has a high correlation of calculated descriptors and experimental results, with a coefficient of determination of 0.82 and a predictive coefficient of determination of 0.92 for unknown molecular structures and conditions. The agreement of calculated molecular interaction orientations with both, experimental results as well as molecular dynamic simulations from literature is shown. The developed descriptors provide the means for improved QSAR models of chromatographic processes, as they reflect the complex interactions of biomolecules with chromatographic phases.