Population serum proteomics uncovers a prognostic protein classifier for metabolic syndrome
Xue Cai, Zhangzhi Xue, Fangfang Zeng, Jun Tang, Liang Yue, Bo Wang, Weigang Ge, Yuting Xie, Zelei Miao, Wanglong Gou, Yuanqing Fu, Sainan Li, Jinlong Gao, Menglei Shuai, Ke Zhang, Fengzhe Xu, Yunyi Tian, Nan Xiang, Yan Zhou, Peng‐Fei Shan, Yi Zhu, Yu‐Ming Chen, Ju‐Sheng Zheng, Tiannan Guo
Abstract
Metabolic syndrome (MetS) is a complex metabolic disorder with a global prevalence of 20%-25%. Early identification and intervention would help minimize the global burden on healthcare systems. Here, we measured over 400 proteins from ∼20,000 proteomes using data-independent acquisition mass spectrometry for 7,890 serum samples from a longitudinal cohort of 3,840 participants with two follow-up time points over 10 years. We then built a machine-learning model for predicting the risk of developing MetS within 10 years. Our model, composed of 11 proteins and the age of the individuals, achieved an area under the curve of 0.774 in the validation cohort (n = 242). Using linear mixed models, we found that apolipoproteins, immune-related proteins, and coagulation-related proteins best correlated with MetS development. This population-scale proteomics study broadens our understanding of MetS and may guide the development of prevention and targeted therapies for MetS.