Quantifying techno-functional properties of ingredients from multiple crops using machine learning.

Food ingredients with a low degree of refining consist of multiple components. Therefore, it is essential to formulate food products based on techno-functional properties rather than composition. We assessed the potential of quantifying techno-functional properties of ingredient blends from multiple crops as opposed to single crops. The properties quantified were gelation, viscosity, emulsion stability, and foaming capacity of ingredients from yellow pea and lupine seeds. The relationships were quantified using spline regression, random forest, and neural networks. Suitable models were picked based on model accuracy and physical feasibility of model predictions. A single model to quantify the properties of both crops could be created for each techno-functional property, albeit with a trade-off of higher prediction errors as compared to models based on individual crops. A reflection on the number of observations in each dataset showed that they could be reduced for some properties.

Functionality-driven food product formulation - An illustration on selecting sustainable ingredients building viscosity.

Currently, food industries typically favour formulation of food products using highly refined techno-functional ingredients of high purity. However, there is a growing interest in less pure techno-functional ingredients with a lower degree of refining as they deliver the same functional properties with reduced environmental impact. We propose that instead of selecting formulations based on purity, they should be selected based on their techno-functional properties. This article illustrates that the shift in perspective may increase the sustainability of food production. The functionality-driven product formulation is explored through a case study in which yellow pea ingredients are selected to increase the viscosity of a salad dressing. The relation between the ingredients (in terms of composition; protein, starch fibre, and a residual fraction) and the final viscosity was quantified and validated using multiple linear regression. The model described the observations well: the final viscosity is mostly dominated by the starch content; protein content has only a marginal impact; and dietary fibre contributes to viscosity with an antagonistic effect with starch. Based on the multiple linear regression model and further formulation optimisation, we identified various combinations of ingredients (with either a high or low degree of refining) that would result in the target final viscosity. An evaluation of the global warming potential of all blends showed that the desired viscosity could be achieved using only isolates, as well as by using only mildly refined fractions. The latter is associated with a global warming potential that is 80% lower than the one based on isolates. This case study demonstrates the proof of concept for this approach, showing it can aid in identifying alternative product formulations with similar techno-functional properties but with a higher sustainability.

Structure analysis of multi-component pastes - The effect of water distribution on the rheological properties.

Techno-functional properties of multi-component blends and ingredients are determined by the contribution of each ingredient and the water distribution between those ingredients in the blends. However, ingredients can consist of multiple components, which should be considered to better understand the properties of ingredients and blends thereof. Recently, empirical models were used to describe the viscosity of mildly refined ingredient blends. While many compositions were described well by the empirical models, blends with high fiber contents were not predicted sufficiently well. Therefore, in this research, the multi-component blends of commercial pea protein, pea starch, and pea fiber isolates were investigated on their rheological properties as a function of dry matter content. The same properties were then measured for blends of two of these isolates mixed in different ratios. From the rheological experiments, estimations of the water distribution were made with the polymer blending law. The results were compared with CLSM images. A quantitative analysis of the CLSM images mostly confirmed the model outcomes. The isolate ratio could describe the isolate blends sufficiently well, meaning that it was not necessary to know the exact compositions of the ingredients. It was concluded that changes in meso-structure of the blends, for example a phase transition at high fiber contents, caused the lower predictability by the recently published empirical viscosity models. This study demonstrates that the water distribution in multi-component blends plays a crucial role for their viscoelastic properties and the contribution of the individual isolates and components. Moreover, these polymer blending laws that include water distribution provide extra mechanical insights into the fraction behavior in multi-component blends.

Investigating the effect of powder manufacturing and reconstitution on casein micelles using asymmetric flow field-flow fractionation (AF4) and transmission electron microscopy.

Milk powders are commonly used for a variety of food products in which among others the milk proteins add to the properties of the products. Processing of milk can, depending on the processing parameters, change the size and structure of the proteins. These changes can be difficult to measure due to the polydispersity of milk components, which makes it a challenge to obtain direct information about the individual proteins. In this paper, the results from an investigation of casein micelle size,size distribution, and structure in reconstituted skim milk and the comparison with raw and pasteurized skim milk are reported. The investigation used asymmetrical flow field-flow fractionation (AF4) in combination with online UV, multi-angle light scattering (MALS), and refractive index (RI) detection and the results were confirmed by transmission electron microscopy (TEM). The results show that there is a difference in casein micelle size distribution between the differently processed milk samples. The casein micelles of the reconstituted milk were found to have a z-average radius of gyration of 72 nm and the casein micelles in the raw and pasteurized skim milk were 58 and 62 nm respectively. The AF4 and TEM data suggest that the cause of the larger casein micelle size is a layer of aggregated whey proteins associated with the casein micelles surface. Moreover, the TEM investigation showed that a larger proportion of the casein micelles are aggregated in reconstituted milk compared to raw and fresh skim milk. Investigation of the effect of reconstitution time shows that the amount of aggregated casein micelles decreases during the first 20 min of reconstitution. The results show that the AF4-method can provide detailed insights into the reconstitution process and properties of different milk samples. Hence, it can be used as a reference or validation for more indirect methods to track the reconstitution of milk powders.

Revisiting the dynamics of proteins during milk powder hydration using asymmetric flow field-flow fractionation (AF4).

The dynamics of ß-casein and casein micelles in the reconstitution of skim milk were revisited in this study. ß-casein migrates into casein micelles upon an increase in temperatures due to an increase in the hydrophobic effect and lower calcium-phosphate cluster solubility. This process can be reversed upon cooling. These phenomena are well known in fresh milk and are not yet clearly established for reconstituted milk powder. As milk powder is commonly used as a functional ingredient in food products, it is of interest to investigate the migration of casein micelle ß-casein to and from the serum phase in reconstituted milk. This study aimed to use asymmetrical flow field flow fractionation (AF4) in combination with various detectors to revisit the dynamics of ß-casein when reconstituting skim milk at different temperatures. Fluorescence-labelled ß-casein was added to fresh and reconstituted skim milk and rapid transport of ß-casein into the outer shell of the casein micelles could be observed already after 5 â€‹min of reconstitution at 50 â€‹°C. This process stabilized after approximately 5 â€‹h, which indicates that an equilibrium of ß-casein between the serum and the micellar phase was reached. Similar results were found for fresh milk. The apparent density of the casein micelles in the skim milk samples was also found to increase during reconstitution at 50 â€‹°C. During cold reconstitution of milk powders, the migration of ß-casein to the serum was not observed. The results suggest that ß-casein was already present in the serum phase upon reconstitution at 6 â€‹°C. When a sample was reconstituted for 180 â€‹min at 50 â€‹°C, the migration of ß-casein back into the serum was observed upon cooling the same sample to 6 â€‹°C. The size of casein micelles in reconstituted milk at 6 â€‹°C was larger compared to reconstitution at 50 â€‹°C. With AF4 and the multi-detector approach, the change in concentration and size of casein micelles can be readily investigated and the migration of ß-casein can be tracked simultaneously. Therefore, the method is a valuable tool for studies of the properties and changes in various milk samples.