A perspective-driven and technical evaluation of machine learning in bioreactor scale-up: A case-study for potential model developments.

Bioreactor scale-up and scale-down have always been a topical issue for the biopharmaceutical industry and despite considerable effort, the identification of a fail-safe strategy for bioprocess development across scales remains a challenge. With the ubiquitous growth of digital transformation technologies, new scaling methods based on computer models may enable more effective scaling. This study aimed to evaluate the potential application of machine learning (ML) algorithms for bioreactor scale-up, with a specific focus on the prediction of scaling parameters. Factors critical to the development of such models were identified and data for bioreactor scale-up studies involving CHO cell-generated mAb products collated from the literature and public sources for the development of unsupervised and supervised ML models. Comparison of bioreactor performance across scales identified similarities between the different processes and primary differences between small- and large-scale bioreactors. A series of three case studies were developed to assess the relationship between cell growth and scale-sensitive bioreactor features. An embedding layer improved the capability of artificial neural network models to predict cell growth at a large-scale, as this approach captured similarities between the processes. Further models constructed to predict scaling parameters demonstrated how ML models may be applied to assist the scaling process. The development of data sets that include more characterization data with greater variability under different gassing and agitation regimes will also assist the future development of ML tools for bioreactor scaling.

Modulation of cream cheese physicochemical and functional properties with ultrafiltration and calcium reduction.

The production of cream cheese from ultrafiltered (UF) milk can reduce acid whey generation but the effect of altered protein and calcium concentration on the physicochemical properties of cream cheese is not well understood. In this study, the effect of skim milk concentration by UF (2.5 and 5 fold) was assessed both with and without calcium reduction using 2% (w/v) cation resin treatment. UF concentration increased the concentration of peptides and free amino acids and led to a more heterogeneous and porous microstructure, resulting in a softer, less viscous and less thermally stable cream cheese. Calcium reduction decreased peptide generation, increased the size of corpuscular structures, decreased porosity and increased thermal stability but did not significantly decrease cheese hardness or viscosity. The study illustrates how protein or calcium concentration, can be used to alter functional properties.

Application of mechanistic modelling in membrane and fiber chromatography for purification of biotherapeutics - A review.

Mechanistic modelling is a simulation tool which has been effectively applied in downstream bioprocessing to model resin chromatography. Membrane and fiber chromatography are newer approaches that offer higher rates of mass transfer and consequently higher flow rates and reduced processing times. This review describes the key considerations in the development of mechanistic models for these unit operations. Mass transfer is less complex than in resin columns, but internal housing volumes can make modelling difficult, particularly for laboratory-scale devices. Flow paths are often non-linear and the dead volume is often a larger fraction of the overall volume, which may require more complex hydrodynamic models to capture residence time distributions accurately. In this respect, the combination of computational fluid dynamics with appropriate protein binding models is emerging as an ideal approach.

Influence of yeast growth conditions and proteolytic enzymes on the amino acid profiles of yeast hydrolysates: Implications for taste and nutrition.

This study investigated the effects of aerobic and anaerobic growth and proteolytic enzymes on the amino acid content of yeast hydrolysates in relation to taste and nutrition. Saccharomyces cerevisiae ATCC5574 was grown under fed-batch aerobic or batch anaerobic conditions. Intracellular glutamic acid (Glu) concentrations were 18-fold higher in aerobic yeast. Hydrolysis with papain and alkaline protease released more amino acids (AA) than simple autolysis or hydrolysis with bromelain, most significantly when applied to aerobic yeast (∼2-fold increase). Autolysates and bromelain hydrolysates from aerobic yeast had low levels of bitter and essential AAs, with high levels of umami Glu. Papain and alkaline protease hydrolysates of aerobic yeast had high levels of umami, bitter and essential AAs. Autolysates/hydrolysates from anaerobic yeast had moderate, high, and low levels of bitter, essential and umami AAs. Selection of both yeast growth conditions and hydrolysis enzyme can manipulate the free AA profile and yield of hydrolysates.

Enhancing the textural and rheological properties of fermentation-induced pea protein emulsion gels with transglutaminase.

The aim of this study was to assess how transglutaminase (TG) impacts the microstructure, texture, and rheological properties of fermentation-induced pea protein emulsion gels. Additionally, the study examined the influence of storage time on the functional properties of these gels. Fermentation-induced pea protein gels were produced in the presence or absence of TG and stored for 1, 4, 8, 12, and 16 weeks. Texture analysis, rheological measurements, moisture content and microstructure evaluation with confocal laser scanning microscopy (CLSM) and 3D image analysis were conducted to explore the effects of TG on the structural and rheological properties of the fermented samples. The porosity of the protein networks in the pea gels decreased in the presence of TG, the storage modulus increased and the textural characteristics were significantly improved, resulting in harder and more springy gels. The gel porosity increased in gels with and without TG after storage but the effect of storage on textural and rheological properties was limited, indicating limited structural rearrangement once the fermentation-induced pea protein emulsion gels are formed. Greater coalescence was observed for oil droplets within the gel matrix after 16 weeks of storage in the absence of TG, consistent with these protein structures being weaker than the more structurally stable TG-treated gels. This study shows that TG treatment is a powerful tool to enhance the textural and rheological properties of fermentation-induced pea protein emulsion gels.

The effect of calcium removal from skim milk by ion exchange on the properties of the ultrafiltration retentate.

New processes are needed to produce concentrated milk feedstocks with tailored calcium content, due to the direct link between calcium concentration and final product texture and functionality. Skim milk treatment with cation exchange resin 1% (w/v) or 2% (w/v) prior to ultrafiltration to a volumetric concentration factor (VCF) of 2.5 or 5 successfully decreased the calcium concentration by 20-30% and produced concentrates with solids content at ∼22-24 g 100 g-1 at a VCF of 5. Calcium reduction partially solubilized the casein micelles, increasing the concentration of soluble protein and individual caseins, leading to decreased turbidity but increased protein hydration and hydrophobicity. Decalcification (2% (w/v) resin treatment) reduced thermal stability, significantly decreasing the denaturation temperature of α-lactalbumin and ß-lactoglobulin in the milk by ∼3 °C and ∼1 °C respectively. Filtration was also altered, reducing permeation flux and the gel concentration and increased filtration time. When combined, calcium reduction and filtration altered functional properties including soluble calcium, soluble protein and sedimentable solids, with increased milk protein hydration also contributing to increased viscosity. This study provides a route to produce calcium-reduced milk concentrates with potential for use in retentate-based dairy products with tailored functionality.

Nanofood Process Technology: Insights on How Sustainability Informs Process Design.

Nanostructured products are an actively growing area for food research, but there is little information on the sustainability of processes used to make these products. In this Review, we advocate for selection of sustainable process technologies during initial stages of laboratory-scale developments of nanofoods. We show that selection is assisted by predictive sustainability assessment(s) based on conventional technologies, including exploratory ex ante and "anticipatory" life-cycle assessment. We demonstrate that sustainability assessments for conventional food process technologies can be leveraged to design nanofood process concepts and technologies. We critically review emerging nanostructured food products including encapsulated bioactive molecules and processes used to structure these foods at laboratory, pilot, and industrial scales. We apply a rational method via learning lessons from sustainability of unit operations in conventional food processing and critically apportioned lessons between emerging and conventional approaches. We conclude that this method provides a quantitative means to incorporate sustainability during process design for nanostructured foods. Findings will be of interest and benefit to a range of food researchers, engineers, and manufacturers of process equipment.

An industrially applicable Escherichia coli platform for bioconversion of thebaine to oripavine and codeine to morphine.

A whole cell Escherichia coli biotransformation platform converting thebaine to oripavine and codeine to morphine was demonstrated with industrially applicable yields (∼1.2 × 10-2 g L-1 h-1 or ∼1.2 × 10-1 g L-1 h-1), improving >13 400-fold upon morphine production in yeast. Mutations enhanced enzyme performance and the use of a purified substrate with rich raw poppy extract expanded applicability.

Industrial freezing and tempering for optimal functional properties in thawed Mozzarella cheese.

Mozzarella cheese was industrially frozen (-18 °C), stored for up to six months, tempered at 4 °C for one or three weeks and the structure and functionality compared to cheese stored at 4 °C and cheese aged at 4 °C for four weeks prior to freezing. When combined with ageing or tempering, the slow industrial freezing minimised changes to the protein network as detected by confocal microscopy and arrested proteolysis. Cheese functionality improved with three weeks of tempering, with properties similar to cheese refrigerated for one month, potentially due to increased proteolysis and protein rehydration. Frozen storage induced ß-sheet and ß-turn structures, as detected by S-FTIR microspectroscopy, with longer tempering leading to structural recovery in the cheese. This study indicates the proteolysis and functionality of frozen cheese can be optimised with tempering time. It also provides new insights into heat transfer during the industrial freezing and tempering of cheese.

Atypical phase behaviour of quinoa protein isolate in mixture with maltodextrin.

Consumers are increasingly looking for new plant-based alternatives to substitute animal proteins in their diets but for some applications it can be difficult to achieve the desired product microstructure using only plant proteins. One approach to facilitate structuring is to mix these plant-based ingredients with a polysaccharide. Here, the phase behaviour and microstructure of quinoa protein isolate (QPI) in mixture with maltodextrin (MD) of two dextrose equivalents (DE 7 and 2) were investigated. The binodals of both QPI-MD phase diagrams showed an atypical shape, where the concentration of MD in the QPI-rich phase and of QPI in the MD-rich phase increased with overall biopolymer concentration. Molecular weight distribution and microstructure analyses revealed that both maltodextrins fractionated between the phases and were probably entrapped within the volume-spanning protein network in the QPI-rich phase, indicating a depletion flocculation mechanism of phase separation. The pre-heating of QPI and the removal of salt from the systems resulted in similarly atypical phase diagrams. The approach presented contributes to our understanding of the phase behaviour of mixtures between plant proteins and polysaccharides, while the results suggest that the formulation of plant-based products of predictable properties may be more challenging than anticipated.

Digital technologies to assess yoghurt quality traits and consumers acceptability.

BACKGROUND: Sensory biometrics provide advantages for consumer tasting by quantifying physiological changes and the emotional response from participants, removing variability associated with self-reported responses. The present study aimed to measure consumers' emotional and physiological responses towards different commercial yoghurts, including dairy and plant-based yoghurts. The physiochemical properties of these products were also measured and linked with consumer responses. RESULTS: Six samples (Control, Coconut, Soy, Berry, Cookies and Drinkable) were evaluated for overall liking by n = 62 consumers using a nine-point hedonic scale. Videos from participants were recorded using the Bio-Sensory application during tasting to assess emotions and heart rate. Physicochemical parameters Brix, pH, density, color (L, a and b), firmness and near-infrared (NIR) spectroscopy were also measured. Principal component analysis and a correlation matrix were used to assess relationships between the measured parameters. Heart rate was positively related to firmness, yaw head movement and overall liking, which were further associated with the Cookies sample. Two machine learning regression models were developed using (i) NIR absorbance values as inputs to predict the physicochemical parameters (Model 1) and (ii) the outputs from Model 1 as inputs to predict consumers overall liking (Model 2). Both models presented very high accuracy (Model 1: R = 0.98; Model 2: R = 0.99). CONCLUSION: The presented methods were shown to be highly accurate and reliable with respect to their potential use by the industry to assess yoghurt quality traits and acceptability. © 2022 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

New Insights into Cheese Microstructure.

Microscopy is often used to assist the development of cheese products, but manufacturers can benefit from a much broader application of these techniques to assess structure formation during processing and structural changes during storage. Microscopy can be used to benchmark processes, optimize process variables, and identify critical control points for process control. Microscopy can also assist the reverse engineering of desired product properties and help troubleshoot production problems to improve cheese quality. This approach can be extended using quantitative analysis, which enables further comparisons between structural features and functional measures used within industry, such as cheese meltability, shreddability, and stretchability, potentially allowing prediction and control of these properties. This review covers advances in the analysis of cheese microstructure, including new techniques, and outlines how these can be applied to understand and improve cheese manufacture.

Plant and Dairy-Based Yogurts: A Comparison of Consumer Sensory Acceptability Linked to Textural Analysis.

Yogurt, readily available in plant and dairy-based formulations, is widely consumed and linked with health benefits. This research is aimed to understand the sensory and textural spectrum of commercially available dairy and plant-based yogurts. In a preliminary study, qualitative focus group discussions (4 groups; n = 32) were used to determine perceptions of 28 dairy and plant-based yogurts, identifying positive consumer perceptions of plant-based yogurts. A smaller subset of five spoonable and one drinkable yogurts-(Reference, Soy, Coconut, Cookies, Berry, and Drinkable) was subsequently selected for rheological and structural measurements, showing wide variations in the microstructure and rheology of selected yogurt samples. A quantitative blind sensory tasting (n = 117) showed varying yogurt acceptability, with Berry being the least-liked and Cookies being the most-liked yogurt, in terms of overall liking. The multi-factor analysis confirmed that compositional and textural elements, including protein content, gel firmness, and consistency coefficient, displayed a positive relationship with overall liking. In contrast, fat, sugar, and calories were negatively correlated to the overall liking. This research showed that texture and other compositional factors are significant determinants of the consumer acceptability of yogurt products and are essential properties to consider in product development.

Differences in Hedonic Responses, Facial Expressions and Self-Reported Emotions of Consumers Using Commercial Yogurts: A Cross-Cultural Study.

Hedonic scale testing is a well-accepted methodology for assessing consumer perceptions but is compromised by variation in voluntary responses between cultures. Check-all-that-apply (CATA) methods using emotion terms or emojis and facial expression recognition (FER) are emerging as more powerful tools for consumer sensory testing as they may offer improved assessment of voluntary and involuntary responses, respectively. Therefore, this experiment compared traditional hedonic scale responses for overall liking to (1) CATA emotions, (2) CATA emojis and (3) FER. The experiment measured voluntary and involuntary responses from 62 participants of Asian (53%) versus Western (47%) origin, who consumed six divergent yogurt formulations (Greek, drinkable, soy, coconut, berry, cookies). The hedonic scales could discriminate between yogurt formulations but could not distinguish between responses across the cultural groups. Aversive responses to formulations were the easiest to characterize for all methods; the hedonic scale was the only method that could not characterize differences in cultural preferences, with CATA emojis displaying the highest level of discrimination. In conclusion, CATA methods, particularly the use of emojis, showed improved characterization of cross-cultural preferences of yogurt formulations compared to hedonic scales and FER.

Myosin sensitivity to thermal denaturation explains differences in water loss and shrinkage during cooking in muscles of distinct fibre types.

The effect of thermal protein denaturation on the structure and quality of muscles of different fibre types is not well understood. Unaged masseter (100% type I fibres) and cutaneous trunci (93% type II fibres) muscles (N = 10) were assessed for their characteristics, protein denaturation, cooking loss, Warner- Bratzler shear force (WBSF) and shrinkage after heating at 50 °C - 85 °C with a rate of 5 °C/ min. Raw masseter had a higher pH, collagen and water content, shorter sarcomere, comparable fibre diameter, and shorter and wider fragments upon homogenization, than cutaneous trunci. In cutaneous trunci, at 55 °C - 60 °C, the lower transition temperature of myosin and the greater cumulative enthalpy resulted in greater cooking loss in muscle cuboids, and greater transverse, longitudinal and volume shrinkage in fibres and fibre fragments, than in masseter. Protein denaturation explained 71% variability in fibre fragment volume and 58% in cooking loss of both muscles, as well as 47% variability in WBSF of masseter.

Ageing and cathepsin inhibition affect the shrinkage of fibre fragments of bovine semitendinosus, biceps femoris and psoas major during heating.

This study examines the effects of ageing (1, 14 days), cathepsin inhibition (No or Yes) and temperature (25-90 °C) on the shrinkage of fibre fragments from three bovine muscles (semitendinosus, biceps femoris and psoas major) during heating. Shrinkage was quantified using light microscopy images. Muscle fibres (except in psoas major) had greater transverse shrinkage, and less longitudinal shrinkage in aged than in unaged muscles at temperatures ≥60-75 °C. In addition, cathepsin inhibition during heating at ≥65-90 °C caused greater transverse shrinkage in semitendinosus fibres, and reduced longitudinal shrinkage for all muscles. At temperatures ≥75 °C, the longitudinal and transverse shrinkage of the fibres was correlated for all muscles. Ageing of biceps femoris increases volume shrinkage on a fibre level, and hence potentially cooking loss, while cathepsin activity in the semitendinosus reduces volume shrinkage. In conclusion, cathepsin activity and ageing influence the shrinkage that occurs during heating and these factors should be explored further to enable optimisation of thermal meat processing.

The effect of pH on the fat and protein within cream cheese and their influence on textural and rheological properties.

The effect of variation in acid gel pH during cream cheese production was investigated. The gel microstructure was denser and cheese texture firmer, as the pH decreased from pH 5.0 to pH 4.3, despite the viscoelasticity of these gels remaining similar during heating. Protein hydration and secondary structure appeared to be key factors affecting both cheese microstructure and properties. Proteins within the matrix appeared to swell at pH 5.0, leading to a larger corpuscular structure; greater ß-turn structure was also observed by synchrotron-Fourier transform infrared (S-FTIR) microspectroscopy and the cheese was softer. A decrease in pH led to a denser microstructure with increased aggregated ß-sheet structure and a firmer cheese. The higher whey protein loss at low pH likely contributed to increased cheese hardness. In summary, controlling the pH of acid gel is important, as this parameter affects proteins in the cheese, their secondary structure and the resulting cream cheese.

Cytochrome P450-mediated N-demethylation of noscapine by whole-cell biotransformation: process limitations and strategies for optimisation.

Cytochrome P450 enzymes catalyse reactions of significant industrial interest but are underutilised in large-scale bioprocesses due to enzyme stability, cofactor requirements and the poor aqueous solubility and microbial toxicity of typical substrates and products. In this work, we investigate the potential for preparative-scale N-demethylation of the opium poppy alkaloid noscapine by a P450BM3 (CYP102A1) mutant enzyme in a whole-cell biotransformation system. We identify and address several common limitations of whole-cell P450 biotransformations using this model N-demethylation process. Mass transfer into Escherichia coli cells was found to be a major limitation of biotransformation rate and an alternative Gram-positive expression host Bacillus megaterium provided a 25-fold improvement in specific initial rate. Two methods were investigated to address poor substrate solubility. First, a biphasic biotransformation system was developed by systematic selection of potentially biocompatible solvents and in silico solubility modelling using Hansen solubility parameters. The best-performing biphasic system gave a 2.3-fold improvement in final product titre compared to a single-phase system but had slower initial rates of biotransformation due to low substrate concentration in the aqueous phase. The second strategy aimed to improve aqueous substrate solubility using cyclodextrin and hydrophilic polymers. This approach provided a fivefold improvement in initial biotransformation rate and allowed a sixfold increase in final product concentration. Enzyme stability and cell viability were identified as the next parameters requiring optimisation to improve productivity. The approaches used are also applicable to the development of other pharmaceutical P450-mediated biotransformations.

High-Efficiency Biocatalytic Conversion of Thebaine to Codeine.

An enzymatic biosynthesis approach is described for codeine, the most widely used medicinal opiate, providing a more environmentally sustainable alternative to current chemical conversion, with yields and productivity compatible with industrial production. Escherichia coli strains were engineered to express key enzymes from poppy, including the recently discovered neopinone isomerase, producing codeine from thebaine. We show that compartmentalization of these enzymes in different cells is an effective strategy that allows active spatial and temporal control of reactions, increasing yield and volumetric productivity and reducing byproduct generation. Codeine is produced at a yield of 64% and a volumetric productivity of 0.19 g/(L·h), providing the basis for an industrially applicable aqueous whole-cell biotransformation process. This approach could be used to redirect thebaine-rich feedstocks arising from the U.S. reduction of opioid manufacturing quotas or applied to enable total biosynthesis and may have broader applicability to other medicinal plant compounds.