Predicting Protein Functionality with Artificial Neural Networks: Foaming and Emulsifying Properties

Using physicochemical properties of 11 food-related proteins, artificial neural networks (ANN) were developed for predicting foam capacity and stability and the emulsion activity index. The prediction accuracy of ANN was compared to that of principal component regression (PCR) models. ANN had better prediction ability than PCR, especially after cross-validation. For foam stability, PCR did not generate a significant model. ANN and PCR models indicated that fluorescence probe hydrophobicity (measured using cispsrinaric acid) and other properties, such as viscosity, surface tension and net charge were important in predicting foam capacity and stability.     

A Simple Fluorometric Method for Fat-Binding Capacity as an Index of Hydrophobicity of Proteins

A new procedure was established for determining fat-binding capacity of proteins, using a nonpolar fluorescent probe diphenylhexatiene dissolved in corn oil. The amount of bound oil was measured from the fluorescence intensity (FI) of the bottom layer after centrifuging the blended protein-oil-probe mixture. This method classified proteins into two groups, i.e., group 1 which had FI increasing linearly with protein concentration and group 2 which had FI increasing then decreasing. The linearly fitted slope of the FI-protein plot revealed better correlation with anilinonaphthalenesulfonate hydrophobicity than with cis-parinarate hydrophobicity. It also showed a significant correlation with emulsifying activity index.     

Nutrtional Evaluation of Covalently Lysine Enriched Wheat Gluten by Tetrahymena Bioassay

Tetrahymena pyriformis W was used in a bioassay to evaluate the improvement in protein quality of wheat gluten by enrichment using either simple addition or carbodiimide-mediated attachment of 1% lysine, Nε-acetyl lysine or Nε-benzylidene lysine. Pepsin predjgestion was more effective in solubilizing samples for the assay than papain. All of the enriched samples had relative nutritive values (RNV) approximating that of the casein standard, which had a value of 100, compared to the gluten and acid-solubilized gluten controls which had RNV of 54 and 64, respectively. No difference in the RNV between lysine and the N^E-lysine derivatives, or between covalent attachment and free addition was detected.     

Internal Friction of Network-Forming Oxide Glasses in the System P2O5-(GeO2, TeO2, Sb2O3, V2O5)

The internal friction of only network-forming oxide glasses containing a P₂O₅ component, i.e., in the systems P₂O₅-GeO₂, P₂O₅-TeO₂, P₂O₅-Sb₂O₃, and P₂O₅-V₂O₅, was measured as a function of temperature by a free torsional vibration method. P₂O₅-TeO₂ and P₂O₅-Sb₂O₃ glasses exhibited clearly high-temperature peaks in a plot of internal friction vs temperature in spite of the absence of nonbridging oxygens and network modifiers. Therefore, we conclude that the high-temperature peaks appeared when strong and weak parts coexisted in the network structure.     

Culture Effect on Ripening Characteristics and Rheological Behavior of Cheddar Cheese

Six defined strains of bacteriophage-insensitive Streptococcus cremoris, grown in whey-based starter media, were used over a period of 10 months to produce more than 2 million kg of Cheddar cheese on continuous cheesemaking equipment. Flavor development in this cheese was less than that in cheese made with conventional bulk starter. Proteolysis in the trichloracetic acid (TCA) soluble cheese fraction was less in cheese manufactured using S. cremoris compared to conventional bulk starter. Rheological properties were studied and the force-compression curves related to the age, composition and pH of the cheese. Pattern recognition techniques were used to analyze the multivariate data. The texture was affected by the mechanical process, moisture content and yield point of the cheese. Casein proteolysis, age, culture type and firmness were the most discriminating variables affecting maturity.     

Comparison of Four Methods for Determining Protease Activity in Milk

Four methods (absorbance at 280 nm; the Lowry method; the fluor-escamine method, and the trinitrobenzenesulfonic acid method) for determining hydrolysis of milk proteins were compared. Each method was applied to the trichloracetic acid soluble fraction of milk protein, which had been digested with trypsin for various periods of time. Detectability was measured as the ratio between standard error of estimate and slope calculated from the linear regression analysis of Deming for cases when both variables were subject to error. Although it was nondimensional, the detectability thus calculated was simple and reliable for comparing assay methods which were based on different analytical principles. Detectability as well as the detection limit measured according to Schwerdtfeger showed that, of the methods compared, the fluorescamine method was most reliable and sensitive.