Vibrational spectroscopy and chemometrics to characterize and quantitate trehalose crystallization

Prediction Machines: Applied Machine Learning for Therapeutic Protein Design and Development

The rapid growth in technological advances and quantity of scientific data over the past decade has led to several challenges including data storage and analysis. Accurate models of complex datasets were previously difficult to develop and interpret. However, improvements in machine learning algorithms have since enabled unparalleled classification and prediction capabilities. The application of machine learning can be seen throughout diverse industries due to their ease of use and interpretability. In this review, we describe popular machine learning algorithms and highlight their application in pharmaceutical protein development. Machine learning models have now been applied to better understand the nonlinear concentration dependent viscosity of protein solutions, predict protein oxidation and deamidation rates, classify sub-visible particles and compare the physical stability of proteins. We also applied several machine learning algorithms using previously published data and describe models with improved predictions and classification. The authors hope that this review can be used as a resource to others and encourage continued application of machine learning algorithms to problems in pharmaceutical protein development.

Protein Aggregation in Frozen Trehalose Formulations: Effects of Composition, Cooling Rate, and Storage Temperature

This study was designed to assess the effects of cooling rate, storage temperature, and formulation composition on trehalose phase distribution and protein stability in frozen solutions. The data demonstrate that faster cooling rates (>100°C/min) result in trehalose crystallization and protein aggregation as determined by Fourier Transform Near-Infrared (FT-NIR) spectroscopy and size-exclusion chromatography, respectively. Conversely, at slower cooling rates (≤1°C/min), trehalose remains predominantly amorphous and there is no effect on protein stability. Evaluation of storage temperatures demonstrates that aggregation increases more rapidly at -14°C compared with higher (-8°C) and lower (-20°C) storage temperatures; however, a relatively higher amount of cumulative aggregation was observed at lower (-20°C) temperature compared with higher storage temperatures (-14°C and -8°C). Further evaluation of the effects of formulation composition suggests that the phase distribution of amorphous and crystallized trehalose dihydrate in frozen solutions depends on the ratio of trehalose to mAb. The results identify an optimal range of trehalose-mAb (w/w) ratio, 0.2-2.4, capable of physically stabilizing mAb formulations during long-term frozen storage-even for fast cooled (>100°C/min) formulations.

Crystal face identification by Raman microscopy for assessment of crystal habit of a drug

Crystal habit is one of the key crystallographic characteristics of active pharmaceutical ingredients (APIs), especially those that are poorly soluble. X-ray powder diffraction has commonly been used to assess crystal habit; however, it can only provide macro-information regarding crystal habit for a whole powder sample, not for individual crystals. We describe an approach that uses Raman microscopy for the identification of crystal faces to assess crystal habit at the individual particle level. An antiepileptic agent, phenytoin, was used as the model substance. Phenytoin crystals form a primitive orthorhombic cell. Raman microscopy was used to identify three different patterns of Raman spectra, corresponding to the crystallographic axis that was parallel to the polarization direction of the excitation laser. Thus, a combination of Raman spectra, in which the polarization direction was horizontal and vertical to the morphologically long axis of the crystal, characterized the crystal face. Phenytoin crystals were prepared under various conditions, and the horizontal/vertical combinations of Raman spectra were recorded for individual crystals. The dominantly exposed crystal faces for each condition were identified. This analytical method enables micro-view assessments of crystal habit, which are helpful for identifying the habits of APIs alone and in formulations.

Determination of amorphous/rubbery states in freeze-dried prebiotic sugars using a combined approach of near-infrared spectroscopy and multivariate analysis

Galacto-oligosaccharides (GOS) and lactulose are well-recognized prebiotics widely used in functional food and pharmaceutical products, but there is still a lack of knowledge regarding their physical-chemical properties. In this study, a physical-chemical approach on two GOS of different composition (GOS Cup Oligo H-70® and GOS Biotempo) and lactulose was assessed. Mid infrared and Raman spectra of the freeze-dried sugars allowed their structural characterization in the amorphous state, lactulose, showing the main spectral differences. Freeze-dried sugars were then equilibrated at 4°C at relative humidity (RH) ranging from 11% to 80%. Near-infrared reflectance spectra were registered in each condition in the 900- to 1700-nm region. A principal component analysis (PCA) was performed on the three sugars equilibrated at different RH. In all the three sugars, the groups observed explained more than 95% of the variance and were related with the RH of the samples. According to the loading plots of PC1, the main differences related with RH were observed in the 1380- to 1500-nm region. As the amorphous states are very sensitive to changes in temperature and moisture content, and the moisture content is related with the parameter T-Tg (T: storage temperature; Tg: vitreous transition temperature), an effort was made to determine this parameter directly from the NIR spectra. To this aim, a partial least square model (PLS) was defined. Tg values obtained by differential scanning calorimetry (DSC) were used to calculate the T-Tg values of reference. The model was validated with an independent set of data. The mean of predicted values fitted nicely T-Tg obtained from DSC (correlation=0.966; R²=0.934), thus supporting the use of the PLS model to investigate unknown samples. The stability of amorphous sugars in foods and pharmaceuticals is of practical and economical importance because it affects different quality attributes of foods, including texture, aroma retention and shelf life. Therefore, predicting T-Tg, a parameter that is independent on the sugar investigated, directly from their NIR spectra is of utmost importance to determine the shelf life of food and food-related products and up to our knowledge has never been determined hereto.

Protein-excipient interactions: mechanisms and biophysical characterization applied to protein formulation development

The purpose of this review is to demonstrate the critical importance of understanding protein-excipient interactions as a key step in the rational design of formulations to stabilize and deliver protein-based therapeutic drugs and vaccines. Biophysical methods used to examine various molecular interactions between solutes and protein molecules are discussed with an emphasis on applications to pharmaceutical excipients in terms of their effects on protein stability. Key mechanisms of protein-excipient interactions such as electrostatic and cation-pi interactions, preferential hydration, dispersive forces, and hydrogen bonding are presented in the context of different physical states of the formulation such as frozen liquids, solutions, gels, freeze-dried solids and interfacial phenomenon. An overview of the different classes of pharmaceutical excipients used to formulate and stabilize protein therapeutic drugs is also presented along with the rationale for use in different dosage forms including practical pharmaceutical considerations. The utility of high throughput analytical methodologies to examine protein-excipient interactions is presented in terms of expanding formulation design space and accelerating experimental timelines.