Neuroimaging and Molecular Markers in Alzheimer’s Disease: Data-Driven Meta-Analysis

Research Background

Alzheimer’s Disease (AD) is a chronic neurodegenerative disease characterized mainly by progressive memory loss and cognitive impairment, making it the most common type of dementia. Neuronal loss, one of the primary hallmarks of AD, is closely associated with gray matter atrophy. Structural Magnetic Resonance Imaging (sMRI)-based studies of brain morphology are an important means of screening and in vivo diagnosing AD. Gray Matter Volume (GMV) and Cortical Thickness (CT) are the most commonly used measures based on sMRI images, reflecting pathological changes from different perspectives. However, due to the small sample size in individual studies, a consensus on the standard atrophy map for AD has not been reached.

Research Motivation and Questions

The main motivation of this research is to assess the susceptibility of localized brain atrophy in AD and its biological mechanisms. Previous meta-analytic studies based on literature suffered from publication bias, heterogeneity in analytical procedures, and inconsistent statistical standards. In contrast, data-driven meta-analyses analyze original data from different sites, making it more robust in detecting differences between case and control groups. The authors aim to obtain more systematic and reliable brain change results through a data-driven meta-analysis.

Source

This paper is co-authored by scholars including Xiaopeng Kang, Dawei Wang, and Jiaji Lin. The research involves institutions such as the Institute of Automation of the Chinese Academy of Sciences, the General Hospital of the People’s Liberation Army of China, and Beijing University of Posts and Telecommunications. The findings are published in the 2024 issue of Neuroscience Bulletin (DOI: 10.1007/s12264-024-01218-x) and supported by several national funds and research projects.

Research Process

Data Collection and Preprocessing

The data for this study were sourced from three multicenter datasets: the proprietary Multicenter AD Imaging Dataset (MCAD), the Alzheimer’s Disease Neuroimaging Initiative (ADNI), and the European Dementia Research Project (EDSD). All participants underwent a series of neuropsychological tests and met specific inclusion criteria. Initially, baseline imaging data from 3,168 subjects were collected. After selecting images with significant noise, 3,118 data points were retained for analysis.

Data preprocessing employed Computational Anatomy Toolbox 12 (CAT12) for standard image segmentation, generating gray matter images. The Brainnetome Atlas was used to calculate GMV and CT for 246 and 210 ROIs, respectively. Low-quality images (resolution, noise, bias, etc.) were excluded, and high-quality data were retained for further research.

Statistical Analysis of Atrophy Patterns

The study employed data-driven meta-analysis to compare brain structures across multiple datasets (from 23 sites) and detect differences between AD cases and normal controls (NC), Mild Cognitive Impairment (MCI) and NC, and AD and MCI. Cohen’s d was used to measure the effect size at each site, and random models and inverse variance methods were used to estimate the weight of each site. The summary effect size for each ROI was calculated, along with z-values and corresponding hypothesis tests.

The study also measured the correlation between gray matter characteristics and cognitive scores using Pearson correlation coefficients, further examining the relationship between atrophy patterns and cognitive decline. Multiple validation analyses were conducted to ensure the reliability of the results, including consistency analysis of effect sizes across sites, calculation of ROI features using other brain maps, and comparison validation using 5,000 surrogate maps generated by spatial autocorrelation.

Biological Pathway Analysis

Using gene transcription data and local brain atrophy patterns from the Allen Human Brain Atlas (AHBA), the study applied Partial Least Squares (PLS) models to analyze the relationship between gene expression and brain structural changes. The results revealed significant gene pathways associated with glutamate signaling pathways and cellular stress responses. Gene Set Enrichment Analysis (GSEA) results corroborated these findings, identifying several related biological pathways. These results provide substantial insights into the biological mechanisms of AD.

Results Verification and Reliability Analysis

Multiple validation experiments, including site correlation analysis, analyses based on different datasets, adoption of different computational models, and Bootstrap validation with subpopulations, demonstrated consistently high reliability. Further confirmation with surrogate maps based on spatial autocorrelation validated the consistency of the main analysis results in terms of gene enrichment pathways.

Main Findings

General Atrophy Patterns

Results indicated that, compared to normal controls, AD and MCI exhibited atrophy in most brain regions, with the hippocampus, amygdala, and temporal lobe being the most notable. The extent of atrophy in AD patients was more severe than in MCI patients. The hippocampus and posterior cingulate were the regions most associated with cognition, with atrophy in these areas highly correlated with cognitive decline.

Biological Mechanisms

The study found that glutamate signaling pathways, cellular stress responses, and synaptic structure and function were closely related to brain atrophy. Regional analysis based on PET images of amyloid-beta (Aβ) and glucose metabolism confirmed significant correlations between AD brain atrophy patterns and Aβ and decreased metabolic activity. These results suggest that Aβ accumulation likely participates in the biological processes related to gray matter atrophy. Additionally, the expression patterns of 5-hydroxytryptamine (5-HT) receptors were significantly correlated with changes in atrophy patterns, indicating that 5-HT receptors might play a protective role in AD brain atrophy progression.

Significance and Value

The significant contribution of this paper lies in its systematic examination of whole-brain atrophy patterns in AD patients through a large-sample data-driven meta-analysis, enriching the understanding of AD pathological manifestations. Additionally, the study identifies several important biological mechanisms through multimodal imaging, offering key insights into AD’s pathogenesis, from glutamate signaling and cellular stress response to synaptic function, providing new perspectives and directions for early detection and treatment strategies.

These findings are valuable for future clinical research and lay a solid foundation for more comprehensive studies on AD pathology and mechanisms.