Proteomic signatures reveal stage-specific progression across gastrointestinal–neuropsychiatric comorbidity: a deep learning and multi-state trajectory analysis

作者：Xiaohua Ye, Zhenhe Jin, Tianyu Zhou, Zhijian Zhao, Chunmin Lou, Chengwei Xu, Kexin Ye, Yandong Li, Yufan Wang, Chaohui Yu, Zhe Shen · 发表于：Scientific Reports · 年份：2026 · DOI：10.1038/s41598-026-64711-5 · 研究领域：Tryptophan and brain disorders、Genetic Associations and Epidemiology、Gut microbiota and health

Gastrointestinal and neuropsychiatric disorders frequently co-occur, but their shared plasma-proteomic associations across disease stages remain unclear. We analyzed 2,920 plasma proteins in 52,953 UK Biobank participants across seven gastrointestinal and nine neuropsychiatric outcomes during a median follow-up of 13.3 years. Proteome-wide Cox analyses identified shared proteins associated with outcomes in both systems. In a strictly nested analysis of 45,811 participants without a study endpoint at baseline, training-only screening identified 198 proteins for CoMorbNet feature prioritization. The cross-system comorbidity model achieved an AUC of 0.717 in a nested held-out validation analysis. Compared with the low-score tertile, the high-score tertile was associated with higher hazards (HRs) across all five disease transitions (HRs, 1.54–2.15; all P < 0.001). Shared proteins were enriched in immune-related pathways. These findings identify stage-specific proteomic patterns associated with gastrointestinal-neuropsychiatric comorbidity, but external validation is required before clinical use.