Graph Neural Networks and Their Current Applications in Bioinformatics

Area of Science:

• Bioinformatics
• Deep Learning
• Graph Neural Networks

Background:

• Graph neural networks (GNNs) excel at processing graph-structured data, making them increasingly vital in bioinformatics due to the proliferation of biological network data.
• GNNs offer advanced capabilities for analyzing complex biological networks, driving innovation across various research domains.

Purpose of the Study:

• To systematically survey the advances and applications of GNNs in bioinformatics from multiple perspectives.
• To introduce fundamental GNN models and their underlying principles.
• To categorize and discuss GNN applications in disease prediction, drug discovery, and biomedical imaging using omics data.

Main Methods:

• Introduction of commonly used GNN models and their basic principles.
• Categorization of GNN tasks based on structural information: node classification, link prediction, and graph generation.
• Discussion of GNN applications across disease prediction, drug discovery, and biomedical imaging, considering various omics data types.

Main Results:

• GNNs have demonstrated excellent performance in numerous biological tasks, including node classification, link prediction, and graph generation.
• Significant progress has been made in applying GNNs to disease prediction, drug discovery, and biomedical imaging.
• Current GNN applications leverage diverse omics data for enhanced biological insights.

Conclusions:

• GNNs represent a powerful and versatile tool for addressing complex challenges in bioinformatics research.
• Despite successes, GNNs face ongoing challenges related to low-quality data processing, methodological limitations, and interpretability.
• Future research directions focus on overcoming these challenges to fully realize the potential of GNNs in biological problem-solving.