Affiliation:
1. Department of Medical and Surgical Sciences, Alma Mater University of Bologna, 40126 Bologna, Italy
2. Centre for Biomedical Image Computing and Analytics, University of Pennsylvania, Philadelphia, PA 19104, USA
3. Department of Clinical Pharmacy and Pharmacology, University of Groningen, University Medical Center Groningen, 9700 AB Groningen, The Netherlands
4. Laboratory of Behavioral Neuroscience, National Institute on Aging, Intramural Research Program, Baltimore, MD 21224, USA
Abstract
Dementia poses a substantial global health challenge, warranting an exploration of its intricate pathophysiological mechanisms and potential intervention targets. Leveraging multi-omic technology, this study utilizes data from 2251 participants to construct classification models using lipidomic, gut metabolomic, and cerebrospinal fluid (CSF) proteomic markers to distinguish between the states of cognitive decline, namely, the cognitively unimpaired state, mild cognitive impairment, and dementia. The analysis identifies three CSF proteins (apolipoprotein E, neuronal pentraxin-2, and fatty-acid-binding protein), four lipids (DEDE.18.2, DEDE.20.4, LPC.O.20.1, and LPC.P.18.1), and five serum gut metabolites (Hyodeoxycholic acid, Glycohyodeoxycholic acid, Hippuric acid, Glyceric acid, and Glycodeoxycholic acid) capable of predicting dementia prevalence from cognitively unimpaired participants, achieving Area Under the Curve (AUC) values of 0.879 (95% CI: 0.802–0.956), 0.766 (95% CI: 0.700–0.835), and 0.717 (95% CI: 0.657–0.777), respectively. Furthermore, exclusively three CSF proteins exhibit the potential to predict mild cognitive impairment prevalence from cognitively unimpaired subjects, with an AUC of 0.760 (95% CI: 0.691–0.828). In conclusion, we present novel combinations of lipids, gut metabolites, and CSF proteins that showed discriminative abilities between the states of cognitive decline and underscore the potential of these molecules in elucidating the mechanisms of cognitive decline.