@article{Liu2026, 
author = {Song Liu and Mengjia Sun and Jianwen Gan and Yuanli Li and Huanhuan Qin and Xinran Ji and Hongxia Chen and Rui Liu and Guangnian Zhao and Bingxin Ma},
title = {Integrated metabolomics and artificial intelligence to predict the dietary-derived compound alleviates cognitive impairment by regulating ferroptosis},
year = {2026},
journal = {Food Science and Human Wellness},
volume = {15},
number = {6},
pages = {9250624},
keywords = {Deep learning, Metabolomics, Cognitive impairment, Linarin, Ferroptosis, High-fat diet},
url = {https://www.sciopen.com/article/10.26599/FSHW.2025.9250624},
doi = {10.26599/FSHW.2025.9250624},
abstract = {Long term high-fat diet (HFD) can damage the central nervous system and lead to cognitive impairment (CI). Compound with anti-CI activity and safety was screened from a dietary-derived compound database based on pathological targets identified by metabolomics integrated with deep learning, and validated by established in vivo and in vitro assays. Ferroptosis was found to be highly associated with HFD-induced brain damage in metabolomic studies. Two deep learning models were used to screen for ferroptosis related target arachidonate-5-lipoxygenase (ALOX5) inhibitory activity and safety evaluation, respectively. The trained models screened new potentially active chemicals from approximately 70000 dietary compounds. Linarin was selected from 143 predicted ALOX5 inhibitors because of its high safety, high oral bioavailability, druglike properties, high blood-brain barrier permeability, and novel chemical structure characteristics. The single-dose acute toxicity study demonstrated the safety of linarin. In vivo and in vitro assays further demonstrated that linarin could alleviate CI by reducing ferroptosis in HFD-induced CI neurons, thereby inhibiting ALOX5 and activating the endogenous antioxidant solute carrier family 7, membrane 11/glutathione peroxidase 4 axis. Overall, linarin can be used as a functional food compound for the treatment of CI. The verification of metabolomics and deep learning screening can assist us in predicting potentially active compounds and revealing their pharmacological mechanisms more efficiently and accurately.}
}