Feature importance of rheological data in predicting structural variations of plant-based biopolymer mixes

Abstract: Justification:
Understanding and predicting the behavior of plant proteins in food matrices remains a challenge. This study focuses on using rheological data to understand, predict, and control structural variations in plant-based biopolymer mixes.



Objectives:
The poster will focus on which independent compositional variables and quantitative rheological parameters best capture structural variation among different biopolymer mixtures. The primary objective was to identify the most relevant features within the results of the rheological analyses performed at industrially relevant conditions of high pressure, concentration, and temperature.



Methods:
The study examines the rheological behavior of 250 biopolymer mixtures, comprising different compositions with plant proteins (14-43% pea and fava bean isolates or concentrate), with different polysaccharide fractions (0-13%) and moisture content (41-70%). Processed under high temperature and pressure, they were analysed using a closed cavity rheometer (CCR). A cluster analysis was performed to identify the relationships between composition variations and rheological properties. Visual inspection of individual variables was performed using scatter plots, while multicomponent analysis was carried out by Principal Component Analysis (PCA).



Results:
The cluster analysis unveils patterns based on protein origin. Solubility and water holding capacity (WHC) of the protein powder correlate with rheological parameters in the linear regime. Concentration of protein, polysaccharide, and moisture correlate more strongly with parameters in the non-linear viscoelastic regime (LAOS). These quantitative non-linear parameters were found to decrease the explained variance in PCA analysis compared to linear regime parameters.

Significance of Research:
Understanding the structure formation of polymer mixtures in a molten state is believed to be pivotal for designing extruded protein-based products. The cluster analysis introduced in this work will serve as a step towards utilizing large rheological data sets to predict structure formation. The outcomes will help to better understand how structural variations influence food properties and how these can be controlled during processing.