Mapping the Genetic-Imaging-Clinical Pathway with Applications to Alzheimer’s Disease

Dengdeng Yu, Linbo Wang, Dehan Kong, Hongtu ZhuJournal of the American Statistical Association 2022 · DOI

TL;DR

This paper introduces a novel two-step statistical approach to untangle the complex interplay between genetics, brain imaging (specifically hippocampal shape), and clinical symptoms in Alzheimer's Disease (AD). By analyzing the ADNI dataset, they identify specific associations between hippocampal shape and cognitive decline, offering a more nuanced understanding of AD's progression.

What Problem Does This Paper Address?

Alzheimer's Disease (AD) is a devastating neurodegenerative disorder characterized by progressive memory loss, cognitive decline, and behavioral changes. While we know that the accumulation of amyloid and tau proteins leads to neuronal damage, particularly in the hippocampus, the precise pathways and how genetic factors influence these changes remain complex and not fully understood. Specifically, researchers want to understand how genetic predispositions translate into observable brain changes (like hippocampal shape alterations) that ultimately drive the clinical symptoms of AD. This requires sophisticated statistical methods to handle the high-dimensional nature of genetic data, detailed brain imaging, and clinical scores simultaneously.

Methods

The authors propose a novel two-step statistical approach to map the "genetic-imaging-clinical pathway" for AD.

  1. Screening: The first step involves a "joint screening" procedure. This is crucial because they are dealing with an extremely large number of genetic covariates (over 6 million) and high-dimensional imaging data (hippocampal surface represented by 100x150 matrices). The joint screening aims to efficiently identify relevant genetic and clinical variables that are associated with both hippocampal shape (the imaging exposure) and cognitive deficits (the outcome), while filtering out irrelevant ones. They developed a "blockwise joint screening" method to better handle the correlated nature of genetic data.
  2. Estimation: The second step uses a penalized regression approach to estimate the association between the selected imaging features and cognitive scores, while accounting for the identified relevant genetic and clinical covariates. This step refines the initial screening to provide a more precise estimate of the relationship.

Datasets

The study utilizes data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. This is a well-established, large-scale, longitudinal study that collects clinical, imaging, genetic, and biomarker data from individuals with and without cognitive impairment. The paper specifically uses data from 566 subjects after preprocessing and quality control.

Key Insights and Ideas

  • Hippocampal Shape Matters: The study highlights that specific geometric features of the hippocampus, captured by radial distance measures across its surface, are significantly associated with cognitive deficits in AD. This goes beyond simple volume measurements.
  • Negative Association: The radial distance of pixels on the hippocampal surface is negatively associated with the severity of behavioral deficits (as measured by ADAS-13 score). This suggests that as the hippocampus changes in specific ways (likely indicating atrophy or structural alterations), cognitive impairment becomes more severe.
  • Regional Specificity: The associations are not uniform across the hippocampus. The CA1 and subiculum subregions show stronger associations with cognitive deficits compared to CA2 and CA3 subregions. This points to specific vulnerabilities within the hippocampus in AD progression.
  • Genetics and Imaging Link: The sophisticated statistical framework allows them to link genetic variants to these imaging-clinical associations, suggesting a pathway where genetic factors influence hippocampal structure, which in turn affects cognitive function.

Key Results

  • The radial distance of each pixel on both the left and right hippocampi is negatively associated with the ADAS-13 cognitive score (a measure of behavioral deficits).
  • These associations are strongest in the CA1 and subiculum subregions of the hippocampus.
  • The shape of the left and right hippocampi collectively account for 5.83% and 4.71% of the total variation in behavioral deficits, respectively. These effect sizes are considered significant.
  • Permutation tests confirmed that these R² values are statistically significant (p-values < 0.001).

Limitations and Future Work

  • Unmeasured Confounding: The authors acknowledge that unmeasured confounders, such as amyloid and tau protein levels, exist and can influence the estimated conditional association. They suggest using genetic variants as potential instrumental variables to control for this in future work.
  • Interaction Effects: The current model primarily focuses on direct associations. Future work could incorporate genetic-genetic and imaging-genetic interaction effects, which are known to be important in complex diseases like AD. This would significantly increase computational demands.
  • Missing Data: The current study removes observations with missing data. Developing methods to accommodate missing exposure, confounders, and outcomes would be valuable.
  • Subgroup Analysis: The relatively small sample size in ADNI limits detailed subgroup analyses (e.g., by diagnosis status). Larger datasets would enable more robust subgroup investigations.
  • Longitudinal Data: The study primarily uses baseline imaging and genetic data with a 24-month follow-up for cognitive scores. Incorporating longitudinal cognitive and imaging data could provide a more comprehensive understanding of disease progression.
  • Other Brain Regions: The study focuses on the hippocampus. Future research could extend this framework to include other brain regions implicated in AD.

Summary updated

Abstract

Alzheimer's disease is a progressive form of dementia that results in problems with memory, thinking, and behavior. It often starts with abnormal aggregation and deposition of <i>β</i> amyloid and tau, followed by neuronal damage such as atrophy of the hippocampi, leading to Alzheimers Disease (AD). The aim of this paper is to map the genetic-imaging-clinical pathway for AD in order to delineate the genetically-regulated brain changes that drive disease progression based on the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset. We develop a novel two-step approach to delineate the association between high-dimensional 2D hippocampal surface exposures and the Alzheimers Disease Assessment Scale (ADAS) cognitive score, while taking into account the ultra-high dimensional clinical and genetic covariates at baseline. Analysis results suggest that the radial distance of each pixel of both hippocampi is negatively associated with the severity of behavioral deficits conditional on observed clinical and genetic covariates. These associations are stronger in Cornu Ammonis region 1 (CA1) and subiculum subregions compared to Cornu Ammonis region 2 (CA2) and Cornu Ammonis region 3 (CA3) subregions. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.
BibTeX
@article{yu2022mapping,
  author = {Dengdeng Yu and Linbo Wang and Dehan Kong and Hongtu Zhu},
  title = {{Mapping the Genetic-Imaging-Clinical Pathway with Applications to Alzheimer’s Disease}},
  journal = {Journal of the American Statistical Association},
  year = {2022},
  doi = {10.1080/01621459.2022.2087658},
}
Figure