Artificial Intelligence-Driven Analysis of Enzymatically Modified Lysozyme: Predicting Bioactivity and Allergenicity of Functional Peptide Fractions

Introduction:
Lysozyme is a well-established antimicrobial protein, commonly used in food preservation due to its efficacy against Gram-positive bacteria. However, its limited activity against Gram-negative strains and allergenic potential present challenges for broader food-related applications. In our previous research, we demonstrated that the enzymatic hydrolysis of lysozyme under specific conditions leads to the formation of bioactive peptide fractions with improved surface properties, expanded antimicrobial spectrum, and reduced immunogenicity. Building on these findings, the present study aims to provide a detailed characterization of the generated lysozyme preparations using advanced multivariate statistical analyses. By evaluating the relationships between hydrolysis conditions and functional outcomes—such as peptide content, surface hydrophobicity, antioxidant capacity, and immunoreactivity—this work supports the rational development of novel, sustainable biofunctional ingredients for the food and health sectors.
Methods:
Sixteen lysozyme variants were generated under varying enzyme concentrations, pH, and reaction times. Each preparation was characterized for hydrolytic and antioxidant activity, surface hydrophobicity, peptide/oligomer content, and immunoreactivity using electrophoresis, ELISA, and Western blot. Multivariate data were analyzed using machine learning models including multilayer perceptrons (MLP), random forests, and principal component analysis (PCA) to identify patterns and predictive relationships.
Results:
MLP and random forest models successfully classified lysozyme preparations based on functional outcomes such as antibacterial efficacy and immunogenic potential. PCA revealed strong correlations between peptide fraction content, hydrophobicity increase, and allergenic signal reduction. The most effective preparations, characterized by >80% peptide content and >40% hydrophobicity increase, were obtained at high pepsin ratios (1:125 to 1:100) and low pH. AI-driven modeling predicted optimal reaction conditions for target functionalities with over 90% accuracy.
Conclusions:
This study integrates enzymatic protein engineering with AI-based data analysis to enable rational design of lysozyme-derived antimicrobial peptides. Our findings highlight how predictive models can guide the development of functional food or pharmaceutical ingredients with reduced allergenicity and enhanced bioactivity, demonstrating a novel interdisciplinary approach for protein optimization.