Artificial intelligence in food bioactive peptides screening: Recent advances and future prospects

Background: Food-derived bioactive peptides (FBPs) play a vital role in nutrition and health. Traditional experimental approaches for identifying FBPs are often labor-intensive, time-consuming, and costly. In contrast, computational approaches, for example, virtual screening and molecular dynamics simulations, have their own limitations. Artificial intelligence (AI) technology enables high-throughput screening and analysis of activity mechanisms for FBPs. Ongoing AI research will enhance the in-depth development and application of FBPs. Scope and approach: This review outlines the general process of AI screening for FBPs, including data foundation, molecular feature representation, machine learning and deep learning model construction and training, as well as evaluation and validation. It also summarizes recent research advances in AI screening of FBPs with different bioactivities, discusses current key issues and challenges, and highlights future research directions and trends of FBPs. Key findings and conclusions: Significant advancements have been made in utilizing AI screening methods to identify functional FBPs with anti-inflammatory, antibacterial, antioxidant, flavor-enhancing, and hypotensive properties, while the research on anti-obesity and anti-fatigue peptides is still at a nascent stage. Deep learning has demonstrated clear predictive advantages over traditional machine learning techniques. However, challenges remain when screening for peptides with different biological activities. Moving forward, data augmentation strategies should be developed within food-specific large models, and a universal deep learning framework based on multi-scale chemical space features should be created to predict peptide-target dynamic interactions. A highthroughput screening framework should be established, alongside enhanced research on AI methods for multifunctional properties like anti-obesity and anti-fatigue effects.