Cross-Attention Graph Neural Networks for Inferring Gene Regulatory Networks with Skewed Degree Distribution

TL;DR

The paper tackles gene regulatory network inference from bulk gene expression data in the presence of skewed degree distributions. It introduces XATGRN, a cross-attention graph neural network that combines a Fusion Module with cross-attention and a DUPLEX-based Complex Dual Graph Embedding using a Hermitian Adjacency Matrix to model connectivity and directionality, enabling prediction of edge existence, direction, and type. Across 9 benchmark datasets, XATGRN consistently outperforms state-of-the-art methods, with ablation studies confirming the necessity of both modules. A breast cancer case study demonstrates biological relevance, identifying hub genes validated by literature and suggesting potential therapeutic drugs, underscoring the model's practical impact for understanding regulatory mechanisms and disease targets.

Abstract

Inferencing Gene Regulatory Networks (GRNs) from gene expression data is a pivotal challenge in systems biology, and several innovative computational methods have been introduced. However, most of these studies have not considered the skewed degree distribution of genes. Specifically, some genes may regulate multiple target genes while some genes may be regulated by multiple regulator genes. Such a skewed degree distribution issue significantly complicates the application of directed graph embedding methods. To tackle this issue, we propose the Cross-Attention Complex Dual Graph Embedding Model (XATGRN). Our XATGRN employs a cross-attention mechanism to effectively capture intricate gene interactions from gene expression profiles. Additionally, it uses a Dual Complex Graph Embedding approach to manage the skewed degree distribution, thereby ensuring precise prediction of regulatory relationships and their directionality. Our model consistently outperforms existing state-of-the-art methods across various datasets, underscoring its efficacy in elucidating complex gene regulatory mechanisms. Our codes used in this paper are publicly available at: https://github.com/kikixiong/XATGRN.

Figures (6)

• Figure 1: Overview of the XATGRN method.
• Figure 2: Visualization of the in-degree and out-degree distribution in DREAM5, Human and E.coli datasets
• Figure 3: Average results over 10 times of 5-fold cross-validationfor model performance comparisons across 9 datasets
• Figure 4: Ablation study in DREAM5 Ecoli cold and human COVID datasets
• Figure 5: Hub gene of the breast cancer