Abstract INTRODUCTION Cerebrospinal fluid (CSF) total tau (t-tau) and phosphorylated tau (p-tau) are biomarkers of Alzheimer’s disease (AD), yet much is unknown about AD-associated changes in tau metabolism and tau tangle etiology. METHODS We assessed the variation of t-tau and p-tau explained by 38 previously identified CSF metabolites using linear regression models in middle-age controls from the Wisconsin Alzheimer’s Disease Research Center, and predicted AD/mild cognitive impairment (MCI) vs. an independent set of older controls using metabolites selected by the least absolute shrinkage and selection operator (LASSO). RESULTS The 38 CSF metabolites explained 70.3% and 75.7% of the variance in t-tau and p-tau. Of these, 7 LASSO-selected metabolites improved the prediction ability of AD/MCI vs. older controls (AUC score increased from 0.92 to 0.97 and 0.78 to 0.93) compared to the base model. DISCUSSION These tau-correlated CSF metabolites increase AD/MCI prediction accuracy and may provide insight into tau tangle etiology.

CSF metabolites associate with CSF tau and improve prediction of Alzheimer’s disease status