The current diagnosis of Parkinson's disease (PD) is mainly based on the typical clinical manifestations. However, 60% dopaminergic neurons have died when the typical clinical manifestations occur. Predictive neurobiomarkers may help identify those PD patients having non-motor disorders or in different stage and achieving the aim of early diagnosis. Up to date, few if any neuroimaging techniques have been described useful for non-movement disorders diagnosis in PD patients. Here, we investigated the alteration of metabolites in PD patients in different stage of PD and non-motor symptoms including sleep, gastrointestinal and cognitive dysfunction, by using the 1H-MRS.A total of 48 subjects were included between 2017 and 2019: 37 PD (15 men, age 47-82 years) and 11 healthy people (8 men, age 49-74 years). All participants underwent MRI and multi-voxel 1H-MRS examination within 3 days in admission. Six kinds of metabolites, such as creatine (Cr), N-acetyl aspartate/creatine (NAA/Cr), N-acetyl aspartate/choline (NAA/Cho), choline/creatine (Cho/Cr), lipid/creatine (LL/Cr), and myo-Inositol/creatine ratio (mI/Cr) were tested among the PD group and the control groups. Statistical analyses and correlation analyses were performed by using SPSS. The p < 0.05 was considered statistically significant.Compared late PD group with a control group or early group, higher Cr ratio and lower NAA/Cr ratio were observed in the late PD group (p < 0.05). The mI/Cr in the late PD group was also lower than that in the early PD group (p < 0.05). Regarding the relationship between metabolites and NMS, Cho/Cr was higher in the sleep disorder group, whereas mI/Cr was lower in the gastrointestinal dysfunction group in comparison with the non-symptom groups. Moreover, Cr, Cho/Cr, mI/Cr, and LL/Cr were identified to have higher concentrations in the cognitive group in thalamus.Proton magnetic resonance spectroscopy is an advanced tool to quantify the metabolic changes in PD. Three biomarkers (Cr, NAA/Cr, and mI/Cr) were detected in the late stage of PD, suggesting that these markers might be potential to imply the progression of PD. In addition, subgroups analysis showed that MRS of thalamus is a sensitive region for the detection of cognitive decline in PD, and the alteration of neurochemicals (involving Cr, Cho, mI, and LL) may be promising biomarkers to predict cognitive decline in PD.Copyright © 2022 Guan, Zheng, Lai, Sun, Geng, Zhang, Zhou, Wu, Chen, Yang, zheng, Wang, Chen and Zhang.