Design of Multifunctional Bioactive Peptides for Potential Application in Food Packaging: A Machine-Learning-Integrated Optimization Framework

Multifunctional bioactive peptides with antioxidant, antifungal, and antibacterial properties can be highly effective for the development of active packaging systems, thereby playing a pivotal role in alleviating foodborne diseases. With peptide design being a combinatorial problem, it is an extremely arduous process to screen the entire sequence space and discover peptides with multifunctional properties. In this study, a computational framework that integrates machine learning and optimization for the design of multifunctional bioactive peptides was proposed. Three machine-learning-based binary classification models were developed to screen antioxidant, antifungal, and antibacterial peptides, respectively. Accuracies for models 1, 2, and 3 were found to be 82.37%, 95.49%, and 96.75%, respectively. These models are then incorporated into an optimization framework to design new bioactive peptides with antioxidant, antifungal, and antibacterial properties. Analysis of amino acid compositions for these newly generated peptides revealed that amino acids that are characteristic of antioxidant, antifungal, and antibacterial peptides are present in significant proportions in these newly generated peptide sequences. This shows that the proposed computational framework indeed generates peptides with multifunctional properties, thereby indicating their potential applications in food packaging. A free-to-use executable tool that can expedite the design of such multifunctional bioactive peptides has also been developed.