Extended Data Fig. 3: Sleep Chart delineates U-shaped associations between sleep duration and 107 organ-associated metabolites.

a) Sleep duration was associated with 107 organ-associated plasma metabolites (x-axis) using the GAM model. These organ-associated metabolites were defined in our previous study³. Non-grey coloured points indicate significant associations after Bonferroni correction for multiple comparisons across all available metabolites (two-sided P₁ < 0.05/327 for all available metabolites). The y-axis displays the effective degrees of freedom (EDF), quantifying the extent of nonlinearity for the U-shaped associations. We annotated representative proteins for each organ system. b-e) Representative U-shaped relationship between 4 metabolites or lipid subclasses and sleep duration. The solid curves depict estimated metabolites, while shaded bands represent the 95% confidence interval (CI). f) Metabolite set enrichment analysis (MSEA) was performed using the Over Representation Analysis (ORA) module in MetaboAnalyst (https://www.metaboanalyst.ca/), which identifies biologically meaningful patterns by assessing whether known metabolite sets (e.g., significant metabolites related to sleep duration) are overrepresented among the list of input metabolites. The analysis uses a hypergeometric test to calculate statistical significance, with multiple testing correction applied using the FDR method. Only metabolite sets with two-sided P_FDR < 0.05 were considered significantly enriched. Only small-molecule metabolites, excluding lipid subclasses, that could be mapped in MetaboAnalyst were included. As the background reference, we used 3,694 metabolite and lipid pathways curated in RaMP-DB (integrating KEGG via HMDB, Reactome, and WikiPathways). We organize distinct metabolite sets along the y-axis by linking them to potential sleep-related hypotheses grounded in prior literature, including the sleep-liver axis, which is closely tied to the energy hypothesis⁸⁸ of sleep.