FOOD SCIENCE ›› 2026, Vol. 47 ›› Issue (20): 361-371.doi: 10.7506/spkx1002-6630-20260407-042

• Reviews • Previous Articles    

Research Advances on Artificial Intelligence-Driven Food-Derived Peptide Self-Assembly: From Prediction to Design

ZHU Chenxiao, YUAN Xin, MA Jingwen, LI Jiahao, DONG Mingzhe, MA Lingjun, CHEN Fang, HU Xiaosong, JI Junfu   

  1. 1. National Engineering Research Center for Fruits and Vegetables Processing, College of Food Science and Nutritional Engineering, China Agricultural University, Beijing 100083, China; 2. Sanya Institute, China Agricultural University, Sanya 572025, China
  • Published:2026-11-06

Abstract: Peptide self-assembly has emerged as a powerful strategy for constructing functional biomaterials with broad potential applications in drug delivery, tissue engineering, and food colloids. However, conventional experimental screening and molecular simulation approaches are often limited by vast combinatorial sequence space, high synthesis costs, and insufficient predictive guidance. In recent years, artificial intelligence (AI) has provided a new paradigm for the efficient prediction and design of peptide self-assembly by leveraging its strength in data mining and pattern recognition. This review systematically summarizes the key elements of AI-based peptide self-assembly prediction, with particular emphasis on dataset construction, feature engineering, AI modeling, and prediction outputs. Specifically, it discusses strategies for dataset development based on accumulated experimental data and AI-assisted literature mining, multidimensional feature engineering approaches, and major modeling paradigms, including conventional machine learning, deep learning, and large language models. The review further highlights recent advances in AI-enabled prediction of peptide assembly propensity, secondary structure, nanomorphology, and functional activity. Finally, it identifies three major challenges in the field: data scarcity and bias, inadequate modeling of environmental factors, and the difficulty of cross-scale prediction, and points out future directions in the development of peptide-specific large language models and data and physics co-driven models, with the aim of providing guidance for the intelligent design and industrialization of peptide self-assembled materials.

Key words: food-derived peptide; self-assembly; artificial intelligence; machine learning; deep learning

CLC Number: