Shared Causal Paths underlying Alzheimer’s dementia and Type 2 Diabetes

Zixin Hu, Rong Jiao, Panpan Wang, Yun Zhu, Jinying Zhao, Phil De Jager, David A. Bennett, Jin Li, Momiao XiongScientific 2020 · DOI

TL;DR

This paper introduces a novel causal inference framework for analyzing multi-omic data to uncover shared biological pathways between Alzheimer's Disease (AD) and Type 2 Diabetes Mellitus (T2DM). By moving beyond traditional association studies, the researchers identified specific shared genes, pathways, and molecular mechanisms that link these seemingly disparate conditions, offering new avenues for understanding and potentially treating them.

What Problem Does This Paper Address?

Alzheimer's Disease (AD) and Type 2 Diabetes Mellitus (T2DM) are distinct conditions, with AD primarily affecting the central nervous system and T2DM a metabolic disorder. However, a growing body of epidemiological and genetic evidence suggests a significant link between them. Current research often relies on association analysis to identify shared genetic factors and pathways. While useful, association studies can only show that two variables are related, not whether one causes the other, thus limiting our understanding of the underlying biological mechanisms. This gap hinders the development of effective therapeutic strategies that target the root causes of these diseases. The authors aim to bridge this gap by developing and applying causal inference methods to multi-omic data to uncover the direct causal relationships and shared pathways between AD and T2DM.

Methods

The core of this paper lies in the development of a novel causal inference framework for genetic studies. This framework moves beyond traditional association analysis by employing causal inference methods to infer multilevel omics causal networks. Key components include:

  • Genome-Wide Causation Studies (GWCS): Unlike Genome-Wide Association Studies (GWAS), GWCS directly tests for causation between genetic variants and diseases using Additive Noise Models (ANMs). This allows for the identification of directed relationships.
  • Multilevel Causal Omics Network Construction: The framework integrates various data types, including genotype, DNA methylation, gene expression, environmental factors, and phenotypes, into a single, connected multilevel genotype-disease network. This is achieved through Structural Equation Models (SEMs) and Integer Programming (IP).
  • Causal Path Search Algorithms: To navigate the complex networks, the authors employ graph theoretical approaches, specifically Depth First Search (DFS) and Breadth First Search (BFS) algorithms, to efficiently identify causal paths from genetic variants to diseases.
  • Simulation Studies: The authors validate their proposed methods (specifically, a Structural Equation Model with Integer Programming, SEMIP) against existing methods like WGCNA and SEM using simulated datasets to evaluate their power and false discovery rate.

Datasets

The primary dataset used in this study is the ROSMAP Project dataset. This dataset comprises longitudinal data from older individuals, including:

  • Genotypes
  • DNA methylation profiles
  • RNA-sequencing (gene expression) data
  • Multiple phenotypes and environmental factors
  • Clinical diagnoses for AD and T2DM

The study analyzed data from 432 individuals who had all these multi-omic layers and phenotypes available.

Key Insights and Ideas

  • Causation over Association: The fundamental shift from association to causation is crucial for understanding disease mechanisms. Identifying direct causal links provides a deeper biological insight than mere correlation.
  • Interconnectedness of AD and T2DM: The study strongly supports the idea that AD and T2DM share significant underlying biological pathways, suggesting a more integrated view of these diseases.
  • "Type 3 Diabetes" Hypothesis: The findings provide empirical evidence that AD can be considered a "type 3 diabetes," highlighting the profound metabolic underpinnings of neurodegenerative diseases.
  • Multi-omic Integration is Key: Unraveling complex disease mechanisms requires integrating diverse biological data (genetics, epigenetics, transcriptomics, etc.) into a unified causal framework.
  • Targeting Shared Pathways: Identifying shared causal genes and pathways offers potential therapeutic targets that could simultaneously benefit patients with AD and T2DM.

Key Results

  • Shared Causal Genes and Pathways: The analysis identified 13 shared causal genes and 16 shared causal pathways between AD and T2DM.
  • Molecular Connections: The study found 754 gene expression nodes and 101 gene methylation nodes that were connected to both AD and T2DM within the multi-omics causal networks.
  • Specific Pathway Identification: Several key pathways were highlighted as shared, including:
  • CREBBP, MAPK, and PI3K-AKT pathways: These pathways were implicated in the link between AD and T2DM, with CREBBP identified as a potential common genetic factor.
  • TTC3, FoxO, MAPK, and PI3K-AKT pathways: Further reinforcing the role of MAPK and PI3K-AKT, this finding also pointed to TTC3 and FoxO as important players.
  • Morphine addiction and neuroactive ligand receptor interaction pathways: These were identified as shared pathways, suggesting broader systemic influences.
  • Fatty acid biosynthesis and primary bile acid biosynthesis pathways: These metabolic pathways were also found to be shared.
  • Methylation's Role: The study demonstrated how shared methylated genes (POU3F2, KIF4B, TNSL3) and pathways (dopaminergic synapse, AMPK) contribute to both AD and T2DM.
  • Methodological Superiority: Simulation results showed that their proposed SEMIP method achieved higher power and lower false discovery rates compared to WGCNA and SEM for causal network construction.

Limitations and Future Work

  • Data Limitations: While the ROSMAP dataset is rich, the authors acknowledge that the number of individuals (432) might limit the power to detect very subtle causal relationships, especially for rare variants.
  • Complexity of Causal Networks: The sheer scale and complexity of the inferred multilevel causal networks present a challenge for manual interpretation. The authors developed algorithms to search these networks, but further refinement and visualization tools could be beneficial.
  • Generalizability: While the ROSMAP dataset is valuable, future work could involve validating these findings in independent cohorts to ensure generalizability.
  • Therapeutic Translation: The identified shared pathways are promising targets, but significant further research is needed to translate these findings into effective therapeutic interventions.
  • "Type 3 Diabetes" Nuance: While the "type 3 diabetes" hypothesis is compelling, further research is needed to fully elucidate the extent and mechanisms of this metabolic link to neurodegeneration.

Summary updated

Abstract

Although Alzheimer's disease (AD) is a central nervous system disease and type 2 diabetes MELLITUS (T2DM) is a metabolic disorder, an increasing number of genetic epidemiological studies show clear link between AD and T2DM. The current approach to uncovering the shared pathways between AD and T2DM involves association analysis; however such analyses lack power to discover the mechanisms of the diseases. As an alternative, we developed novel causal inference methods for genetic studies of AD and T2DM and pipelines for systematic multi-omic casual analysis to infer multilevel omics causal networks for the discovery of common paths from genetic variants to AD and T2DM. The proposed pipelines were applied to 448 individuals from the ROSMAP Project. We identified 13 shared causal genes, 16 shared causal pathways between AD and T2DM, and 754 gene expression and 101 gene methylation nodes that were connected to both AD and T2DM in multi-omics causal networks.
BibTeX
@inproceedings{hu2020shared,
  author = {Zixin Hu and Rong Jiao and Panpan Wang and Yun Zhu and Jinying Zhao and Phil De Jager and David A. Bennett and Jin Li and Momiao Xiong},
  title = {{Shared Causal Paths underlying Alzheimer’s dementia and Type 2 Diabetes}},
  booktitle = {Scientific},
  year = {2020},
  doi = {10.1038/s41598-020-60682-3},
}
Figure