Potential non-invasive biomarkers of chronic sleep disorders identified by salivary metabolomic profiling among middle-aged Japanese men

Abstract

Sleep disorders have become a global social problem that increases the risk of developing mental illnesses and metabolic diseases. We aimed to identify biomarkers with which to non-invasively and objectively evaluate chronic sleep disorders. We used capillary electrophoresis-Fourier transform mass spectrometry (CE-FTMS) to analyze metabolomes in saliva collected from 50 persons each with good (≤ 2) and poor (≥ 6) sleep quality scored according to the Japanese version of the Pittsburgh Sleep Quality Index (PSQI-J) self-report questionnaire. The levels of five metabolites including glycerol and hippuric acid and eight including 2-hydroxybutyric acid (2HB), were respectively decreased and increased in participants with poor sleep quality. We established a random forest model consisting of six metabolites, including glycerol and hippuric acid, with a prediction accuracy of 0.866. Correlations between metabolites and sleep satisfaction were assessed using the Oguri-Shirakawa-Azumi sleep inventory, middle-age and aged version (OSA-MA) questionnaire. The results showed that 2'-deoxyguanosine, N1-acetylspermine, and 2,4-dihydroxybenzoic acid correlated positively, whereas glucosamine 6-phosphate and trimethylamine N-oxide correlated negatively with sleep quality. These findings suggested that changes in salivary metabolites reflect pathophysiological mechanisms of chronic sleep disorders, and that saliva samples could serve as non-invasive and objective diagnostic targets for predicting habitual sleep quality.

Keywords: Biomarker; Chronic sleep disorder; Insomnia; Metabolome; Microbiota; Saliva.

Fig. 1

Principal component analysis scores. Blue circles and red triangles indicate groups with good and poor sleep quality, respectively. PCA, principal component analysis.

Fig. 2

Heatmap and hierarchical cluster analysis. The heatmap represents the scaled intensity levels of metabolites across samples, with rows indicating metabolites and columns representing individual participants. Hierarchical cluster analysis (HCA) dendrograms are displayed on the left, showing clustering patterns for metabolites. Positive values are shown in red, negative values in green, and values close to zero are shown in white.

Fig. 3

Partial least squares scores. Blue circles and red triangles indicate groups with good and poor sleep quality, respectively. PLS, partial least squares.

Fig. 4

Receiver operating characteristic curve of random forest models constructed using six metabolites.