Chemometrics in Protein Formulation: Stability Governed by Repulsion and Protein Unfolding.

Therapeutic proteins can be challenging to develop due to their complexity and the requirement of an acceptable formulation to ensure patient safety and efficacy. To date, there is no universal formulation development strategy that can identify optimal formulation conditions for all types of proteins in a fast and reliable manner. In this work, high-throughput characterization, employing a toolbox of five techniques, was performed on 14 structurally different proteins formulated in 6 different buffer conditions and in the presence of 4 different excipients. Multivariate data analysis and chemometrics were used to analyze the data in an unbiased way. First, observed changes in stability were primarily determined by the individual protein. Second, pH and ionic strength are the two most important factors determining the physical stability of proteins, where there exists a significant statistical interaction between protein and pH/ionic strength. Additionally, we developed prediction methods by partial least-squares regression. Colloidal stability indicators are important for prediction of real-time stability, while conformational stability indicators are important for prediction of stability under accelerated stress conditions at 40 °C. In order to predict real-time storage stability, protein-protein repulsion and the initial monomer fraction are the most important properties to monitor.

Water quality monitoring based on chemometric analysis of high-resolution phytoplankton data measured with flow cytometry.

River water is an important source of Dutch drinking water. For this reason, continuous monitoring of river water quality is needed. However, comprehensive chemical analyses with high-resolution gas chromatography [GC]-mass spectrometry [MS]/liquid chromatography [LC]-MS are quite tedious and time consuming; this makes them poorly fit for routine water quality monitoring and, therefore, many pollution events are missed. Phytoplankton are highly sensitive and responsive to toxicity, which makes them highly usable for effect-based water quality monitoring. Flow cytometry can measure the optical properties of phytoplankton every hour, generating a large amount of information-rich data in one year. However, this requires chemometrics, as the resulting fingerprints need to be processed into information about abnormal phytoplankton behaviour. We developed Discriminant Analysis of Multi-Aspect CYtometry (DAMACY) to model the "normal condition" of the phytoplankton community imposed by diurnal, meteorological, and other exogenous influences. DAMACY first describes the cellular variability and distribution of phytoplankton in each measurement using principal component analysis, and then aims to find subtle differences in these phytoplankton distributions that predict normal environmental conditions. Deviations from these normal environmental conditions indicated abnormal phytoplankton behaviour that happened alongside pollution events measured with the GC/MS and LC/MS systems. Thus, our results demonstrate that flow cytometry in combination with chemometrics may be used for an automated hourly assessment of river water quality and as a near real-time early warning for detecting harmful known or unknown contaminants. Finally, both the flow cytometer and the DAMACY algorithm run completely autonomous and only requires maintenance once or twice per year. The warning system results may be uploaded automatically, so that drinking water companies may temporary stop pumping water whenever abnormal phytoplankton behaviour is detected. In the case of prolonged abnormal phytoplankton behaviour, comprehensive analysis may still be used to identify the chemical compound, its origin, and toxicity.

Isothermal Chemical Denaturation: Data Analysis, Error Detection, and Correction by PARAFAC2.

Characterization of a protein's conformational stability is a key step in the development of biotherapeutics, where protein unfolding leads to adverse properties, such as aggregation and loss of efficacy. Isothermal chemical denaturation (ICD) can be applied to determine chemical stability, aiming to identify the optimal solvent conditions, in terms of pH, salt concentration, and added excipients. For seven monoclonal antibodies, this study investigates the observed intrinsic protein fluorescence emission spectra as a function of denaturant concentration. Protein formulations are screened in two experimental series. We show how the peak shapes of folded and unfolded proteins are preserved under added salt (0-140 mM NaCl) and added excipients concentrations, as typically found in biotherapeutic formulations and that only minor effects in tryptophan fluorescence peak tailing are observed over a large pH range (5.5-9.0). The data of seven mAbs, where GuHCl was a suitable denaturant, are modeled using PARAFAC2. PARAFAC2, a linear decomposition method, is well suited for the data and yields robust, valid, and automated models that allow for the detection of erroneous measurements. Analysis of the errors show correlation with the well-based experimental setup, and differences in observed errors between the two experimental series. We additionally show a correction method for these outliers based on PARAFAC2 model scores, such that full transition curves can be retrieved, increasing the accuracy of any subsequent analysis.

Advancing Therapeutic Protein Discovery and Development through Comprehensive Computational and Biophysical Characterization.

Therapeutic protein candidates should exhibit favorable properties that render them suitable to become drugs. Nevertheless, there are no well-established guidelines for the efficient selection of proteinaceous molecules with desired features during early stage development. Such guidelines can emerge only from a large body of published research that employs orthogonal techniques to characterize therapeutic proteins in different formulations. In this work, we share a study on a diverse group of proteins, including their primary sequences, purity data, and computational and biophysical characterization at different pH and ionic strength. We report weak linear correlations between many of the biophysical parameters. We suggest that a stability comparison of diverse therapeutic protein candidates should be based on a computational and biophysical characterization in multiple formulation conditions, as the latter can largely determine whether a protein is above or below a certain stability threshold. We use the presented data set to calculate several stability risk scores obtained with an increasing level of analytical effort and show how they correlate with protein aggregation during storage. Our work highlights the importance of developing combined risk scores that can be used for early stage developability assessment. We suggest that such scores can have high prediction accuracy only when they are based on protein stability characterization in different solution conditions.

Novel non-linear curve fitting to resolve protein unfolding transitions in intrinsic fluorescence differential scanning fluorimetry.

In biotherapeutic protein research, an estimation of the studied protein's thermal stability is one of the important steps that determine developability as a function of solvent conditions. Differential Scanning Fluorimetry (DSF) can be applied to measure thermal stability. Label-free DSF measures amino acid fluorescence as a function of temperature, where conformational changes induce observable peak deformation, yielding apparent melting temperatures. The estimation of the stability parameters can be hindered in the case of multidomain, multimeric or aggregating proteins when multiple transitions partially coincide. These overlapping protein unfolding transitions are hard to evaluate by the conventional methodology, as peak maxima are shifted by convolution. We show how non-linear curve fitting of intrinsic fluorescence DSF can deconvolute highly overlapping transitions in formulation screening in a semi-automated process. The proposed methodology relies on synchronous, constrained fits of the fluorescence intensity, ratio and their derivatives, by combining linear baselines with generalized logistic transition functions. The proposed algorithm is applied to data from three proteins; a single transition, a double separated transition and a double overlapping transition. Extracted thermal stability parameters; apparent melting temperatures Tm,1, Tm,2 and melting onset temperature Tonset are obtained and compared with reference software analysis. The fits show R2 = 0.94 for single and R2 = 0.88 for separated transitions. Obtaining values and trends for Tonset in a well-described and automated way, will aid protein scientist to better evaluate the thermal stability of proteins.

Application of interpretable artificial neural networks to early monoclonal antibodies development.

The development of a new protein drug typically starts with the design, expression and biophysical characterization of many different protein constructs. The initially high number of constructs is radically reduced to a few candidates that exhibit the desired biological and physicochemical properties. This process of protein expression and characterization to find the most promising molecules is both expensive and time-consuming. Consequently, many companies adopt and implement philosophies, e.g. platforms for protein expression and formulation, computational approaches, machine learning, to save resources and facilitate protein drug development. Inspired by this, we propose the use of interpretable artificial neuronal networks (ANNs) to predict biophysical properties of therapeutic monoclonal antibodies i.e. melting temperature Tm, aggregation onset temperature Tagg, interaction parameter kD as a function of pH and salt concentration from the amino acid composition. Our ANNs were trained with typical early-stage screening datasets achieving high prediction accuracy. By only using the amino acid composition, we could keep the ANNs simple which allows for high general applicability, robustness and interpretability. Finally, we propose a novel "knowledge transfer" approach, which can be readily applied due to the simple algorithm design, to understand how our ANNs come to their conclusions.