Imaging Genetic Based Mediation Analysis for Human Cognition

Tingan Chen, Abhishek Mandal, Hongtu Zhu, Rongjie LiuFrontiers in Neuroscience 2022 · DOI

TL;DR

This paper introduces a novel framework for understanding how genetic variations influence human cognition by examining the brain's structural connectome as a mediator. By integrating genetic data, brain imaging, and cognitive assessments, the study identifies specific genetic pathways that impact cognitive traits, offering a more nuanced view of the gene-brain-behavior relationship.

What Problem Does This Paper Address?

Understanding the intricate relationship between our genes, brain structure, and cognitive abilities is a major challenge in neuroscience and genetics. While it's known that genetics influences cognition, the precise mechanisms are often unclear. Specifically, how genetic predispositions translate into observable cognitive differences through the brain's complex network of connections (the connectome) has been difficult to pinpoint. Traditional approaches often struggle with the high dimensionality of genetic data and the complexity of representing brain connectivity. This paper aims to bridge this gap by developing a statistical framework to trace the causal pathways from genetic variants (SNPs) to cognitive traits, with the brain's structural connectome acting as the intermediary.

Methods

The authors propose an "imaging genetic-based mediation analysis" framework. The core methodology involves several key steps:

  1. Brain Connectome Representation: Instead of using traditional region-of-interest (ROI) based atlases, they develop a data-driven representation of the brain connectome. This involves:
  2. Fiber Tracking and Clustering: Diffusion MRI data is used to track white matter fibers. These fibers are then clustered using a mini-batch K-means algorithm to create a set of "population clusters" representing distinct connectivity patterns.
  3. Connectome Vector: For each individual, their connectome is represented as a vector where each element signifies the proportion of their fibers falling into each cluster. This provides a more flexible and data-driven representation compared to fixed atlases.
  4. Genome-Wide Association Study (GWAS) on Connectome: A GWAS is performed to identify Single Nucleotide Polymorphisms (SNPs) that are significantly associated with these connectome representations. This step reduces the dimensionality of the genetic data by focusing on SNPs that have a demonstrable link to brain structure. False Discovery Rate (FDR) control is used for multiple testing correction.
  5. Mediation Analysis: Once significant SNP-connectome associations are found, a mediation analysis is conducted. This statistical approach aims to determine if the effect of an SNP on a cognitive trait is transmitted through the brain connectome. The model estimates:
  6. Direct Effect: The direct influence of an SNP on a cognitive trait, independent of the connectome.
  7. Indirect Effect: The influence of an SNP on a cognitive trait that is mediated by the connectome.
  8. Confounders: The model also accounts for confounders like age and gender.

Datasets

The study utilizes data from the Human Connectome Project (HCP) 1200 Subjects Release. This dataset is comprehensive, providing:

  • Genetic data: Genome-wide SNP data.
  • Diffusion MRI data: Used for constructing the brain connectome.
  • Behavioral/Cognitive data: Measures of various cognitive traits (oral reading score, list sorting score, card sort score, and receptive vocabulary score).
  • Confounder data: Age and gender.

The authors specifically mention using data from 870 individuals after quality control and handling of twin data (removing one twin from MZ pairs to avoid inflating Type I error rates).

Key Insights and Ideas

  • Connectome as a Mediator: The central idea is that the brain's structural connectivity isn't just a passive substrate but an active mediator through which genetic predispositions influence cognitive function.
  • Data-Driven Connectome Representation: Moving away from pre-defined atlases to a data-driven clustering approach for connectome representation offers a more personalized and potentially more sensitive way to capture individual brain structure.
  • Identifying Specific Pathways: The framework allows for the identification of specific SNPs and corresponding brain connectome clusters that mediate cognitive traits, moving beyond general associations.
  • Direct vs. Indirect Effects: The analysis distinguishes between direct genetic effects on cognition and indirect effects mediated by brain structure, providing a more detailed understanding of genetic influence.
  • Moderate Mediation: The study found that a moderate percentage (8-15%) of the global effect of certain SNPs on cognitive traits is mediated by the brain connectome, suggesting that while direct effects exist, the connectome plays a significant role.

Key Results

  • 942 Significant SNPs: The GWAS identified 942 SNPs significantly associated with 5 out of 10 brain connectome clusters.
  • Direct Effects Dominant: A majority of the selected SNPs showed significant direct effects on human cognitive traits.
  • Significant Indirect Effects Found: For some SNPs, significant indirect effects mediated by the brain connectome were also identified for traits like ReadEng and PicVocab. For instance, SNP rs145868200 showed a significant indirect effect on the "PicVocab" trait through the connectome.
  • Trait-Specific Pathways: The analysis revealed that specific SNPs and connectome clusters are associated with particular cognitive traits, highlighting the specificity of these gene-brain-cognition relationships.
  • Example SNP Analysis: The paper highlights SNP rs138291276, which showed both significant direct and indirect effects on PicVocab and ReadEng, and is linked to genes involved in receptor stabilization. Another SNP, rs4888388, showed a significant indirect effect but not a significant direct effect, suggesting its influence is primarily mediated by the connectome.

Limitations and Future Work

  • Twin/Family Study Challenges: The current framework struggles to fully account for the complexities of twin and family studies, where genetic relatedness can introduce biases. The method of removing one twin from MZ pairs, while standard, reduces sample size and data richness.
  • Optimal Cluster Number: Determining the optimal number of clusters for the connectome representation remains an open problem. While 10 clusters were used for better visualization and interpretability, further structural justification is needed.
  • Sample Size and Replication: The authors acknowledge the need for larger sample sizes and external replication of their findings to increase confidence and generalizability.
  • Advanced Statistical Models: Developing more advanced statistical models to handle the nuances of family data and improve the robustness of the analysis is suggested.
  • Leveraging Brain Structure: Future work could explore integrating more detailed brain structure information to better inform the clustering process and identify optimal cluster numbers.

Summary updated

Abstract

The brain connectome maps the structural and functional connectivity that forms an important neurobiological basis for the analysis of human cognitive traits while the genetic predisposition and our cognition ability are frequently found in close association. The issue of how genetic architecture and brain connectome causally affect human behaviors remains unknown. To seek for the potential causal relationship, in this paper, we carried out the causal pathway analysis from single nucleotide polymorphism (SNP) data to four common human cognitive traits, mediated by the brain connectome. Specifically, we selected 942 SNPs that are significantly associated with the brain connectome, and then estimated the direct and indirect effect on the human traits for each SNP. We found out that a majority of the selected SNPs have significant direct effects on human traits and discussed the trait-related brain regions and their implications.
BibTeX
@inproceedings{chen2022imaging,
  author = {Tingan Chen and Abhishek Mandal and Hongtu Zhu and Rongjie Liu},
  title = {{Imaging Genetic Based Mediation Analysis for Human Cognition}},
  booktitle = {Frontiers in Neuroscience},
  year = {2022},
  doi = {10.3389/fnins.2022.824069},
}
Figure