Systematic review of AI applications for protein misfolding diseases

Neurodegenerative diseases encompass a diverse group of disorders ranging from adult-onset conditions such as Alzheimer’s and Parkinson’s disease to pediatric forms including neuronal ceroid lipofuscinoses (NCLs), Niemann-Pick type C (NPC), and infantile neuroaxonal dystrophy (INAD), all of which are characterized by protein misfolding and chronic neuroinflammation. During their occurrence and development, the innate immune system, especially the immune responses mediated by microglia in the central nervous system, plays a crucial regulatory role. Increasing evidence indicates that misfolded and abnormally aggregated proteins, such as β-amyloid (Aβ), Tau, α-synuclein, and TDP-43, are not only neurotoxic factors but can also act as damage-associated molecular patterns (DAMPs) recognized by innate immune receptors, thereby triggering persistent neuroinflammatory responses. However, traditional experimental and computational methods still have significant limitations in systematically analyzing the “protein misfolding–innate immune activation” mechanism. In recent years, artificial intelligence has made breakthrough progress in protein structure prediction, multi-conformation modeling, and integration of multi-omics data, providing a new research paradigm for revealing the intrinsic relationship between protein misfolding and innate immunity across the spectrum of neurodegenerative diseases. This article systematically reviews the latest applications of artificial intelligence in predicting the conformational characteristics of misfolded proteins, simulating the protein aggregation process, revealing the mechanism of innate immune perception, and reconstructing the regulatory network of neuroinflammation. It focuses on discussing the significance of deep learning models such as AlphaFold, I-TASSER, RoseTTAFold, Phyre2, and ESMFold in the field of protein structure prediction, as well as the related research on multi-modal AI technology in revealing the complex molecular mechanisms behind neurodegenerative diseases, such as combining AI with mathematical models to simulate the spread of misfolded proteins and further exploring the association with disease progression. The review also highlights the potential of AI to address the diagnostic challenges unique to pediatric neurodegenerative disorders, which, despite their rarity, collectively impose devastating lifelong burdens. In summary, AI tools not only deepen our understanding of the molecular mechanisms underlying both adult and childhood neurodegenerative diseases but also open up new avenues for developing innovative diagnostic tools and treatment methods.