Abstract
INTRODUCTION: We investigated longitudinal cognitive trajectories in relation to cerebrospinal fluid (CSF) proteomics in Alzheimer's disease (AD). METHODS: Differential protein abundance analysis of proteomics (SomaScan 7K) data was performed for composite scores for domain-specific cognitive functions (memory [MEM], executive function [EF], and language [LAN]) and their longitudinal changes in AD. This was followed by co-abundance network analysis, functional enrichment analysis, and machine learning to identify co-abundant protein modules, associated biological pathways, and prediction models of cognitive trajectories. RESULTS: We identified proteins and pathways associated with composite scores for domain-specific cognitive functions and their longitudinal changes. Proteins related to MEM and EF were primarily enriched in microglia and oligodendrocytes, respectively. The prediction model for the trajectories of cognitive decline achieved an area under the curve of up to 0.79. DISCUSSION: Our findings suggest protein signatures and their functional pathways associated with longitudinal changes for cognitive decline, providing molecular insights into longitudinal cognitive trajectories in AD. HIGHLIGHTS: Altered cerebrospinal fluid (CSF) protein levels were associated with cognitive functions at baseline. Altered CSF protein levels were associated with changes of cognitive decline. Proteins associated with memory were primarily enriched in microglia. Proteins associated with executive function were enriched in oligodendrocytes. The prediction performance for cognitive trajectories was improved up to 30.9%.