Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Heterogeneous Causal Metapath Graph Neural Network for Gene-Microbe-Disease Association Prediction

Kexin Zhang, Feng Huang, Luotao Liu, Zhankun Xiong, Hongyu Zhang, Yuan Quan, Wen Zhang

TL;DR

This work tackles the prediction of triple-wise gene–microbe–disease ($GMD$) associations by introducing Heterogeneous Causal Metapath Graph Neural Network (HCMGNN). HCMGNN builds a $GMD$ heterogeneous graph, derives six directed causal subgraphs via causal metapaths, and learns multi-view embeddings through intra-subgraph causal semantic sharing followed by inter-subgraph attention-based fusion. The model demonstrates superior predictive performance over a range of baselines, with ablations confirming the importance of causal metapaths, shared semantics, and feature information, and shows particular strength for sparse, low-degree triplets. This approach provides a principled framework to capture directional, high-order interactions among genes, microbes, and diseases, potentially guiding experimental validation and accelerating discovery of novel GMD associations.

Abstract

The recent focus on microbes in human medicine highlights their potential role in the genetic framework of diseases. To decode the complex interactions among genes, microbes, and diseases, computational predictions of gene-microbe-disease (GMD) associations are crucial. Existing methods primarily address gene-disease and microbe-disease associations, but the more intricate triple-wise GMD associations remain less explored. In this paper, we propose a Heterogeneous Causal Metapath Graph Neural Network (HCMGNN) to predict GMD associations. HCMGNN constructs a heterogeneous graph linking genes, microbes, and diseases through their pairwise associations, and utilizes six predefined causal metapaths to extract directed causal subgraphs, which facilitate the multi-view analysis of causal relations among three entity types. Within each subgraph, we employ a causal semantic sharing message passing network for node representation learning, coupled with an attentive fusion method to integrate these representations for predicting GMD associations. Our extensive experiments show that HCMGNN effectively predicts GMD associations and addresses association sparsity issue by enhancing the graph's semantics and structure.

Heterogeneous Causal Metapath Graph Neural Network for Gene-Microbe-Disease Association Prediction

TL;DR

This work tackles the prediction of triple-wise gene–microbe–disease () associations by introducing Heterogeneous Causal Metapath Graph Neural Network (HCMGNN). HCMGNN builds a heterogeneous graph, derives six directed causal subgraphs via causal metapaths, and learns multi-view embeddings through intra-subgraph causal semantic sharing followed by inter-subgraph attention-based fusion. The model demonstrates superior predictive performance over a range of baselines, with ablations confirming the importance of causal metapaths, shared semantics, and feature information, and shows particular strength for sparse, low-degree triplets. This approach provides a principled framework to capture directional, high-order interactions among genes, microbes, and diseases, potentially guiding experimental validation and accelerating discovery of novel GMD associations.

Abstract

The recent focus on microbes in human medicine highlights their potential role in the genetic framework of diseases. To decode the complex interactions among genes, microbes, and diseases, computational predictions of gene-microbe-disease (GMD) associations are crucial. Existing methods primarily address gene-disease and microbe-disease associations, but the more intricate triple-wise GMD associations remain less explored. In this paper, we propose a Heterogeneous Causal Metapath Graph Neural Network (HCMGNN) to predict GMD associations. HCMGNN constructs a heterogeneous graph linking genes, microbes, and diseases through their pairwise associations, and utilizes six predefined causal metapaths to extract directed causal subgraphs, which facilitate the multi-view analysis of causal relations among three entity types. Within each subgraph, we employ a causal semantic sharing message passing network for node representation learning, coupled with an attentive fusion method to integrate these representations for predicting GMD associations. Our extensive experiments show that HCMGNN effectively predicts GMD associations and addresses association sparsity issue by enhancing the graph's semantics and structure.
Paper Structure (23 sections, 10 equations, 4 figures, 1 table)

This paper contains 23 sections, 10 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Biomedical Association Mining
  4. Metapath-based HGNNs
  5. Methodology
  6. Problem Formulation
  7. Model Architecture
  8. Feature Transformation
  9. Causal Subgraph Generation
  10. Intra-Subgraph Message Passing
  11. Causal Semantic Encoder
  12. Message Aggregation
  13. Inter-Subgraph Attentive Fusion
  14. Model Training
  15. Results
  16. ...and 8 more sections

Figures (4)

  • Figure 1: Overall architecture of HCMGNN. HCMGNN mainly includes: (1) feature transformation; (2) causal subgraph generation; (3) intra-subgraph message passing; (4) inter-subgraph attentive fusion; (5) model training.
  • Figure 2: The t-SNE visualization of four models.
  • Figure 3: Performances of HCMGNN and variants in ablation study.
  • Figure 4: Performances of HCMGNN and w/o MP in test triplets with different average node degree (N) distributions. Each bar shows the result on the triplets in the average node degree interval (0, N].