The Pathway-Informed Deep Learning Models in Cancer Research: A Survey

Area of Science:

• Computational biology and bioinformatics
• Deep learning applications in oncology

Background:

• Biological pathways are crucial for understanding complex diseases, drug discovery, and personalized medicine.
• Deep learning models incorporating pathway information show promise in cancer research due to interpretability and performance.

Purpose of the Study:

• To systematically review and categorize pathway-informed deep learning models used in cancer research.
• To analyze the strategies for applying pathway information within these models.
• To identify common pathway databases and interpretability methods.

Main Methods:

• A comprehensive survey of pathway-informed deep learning models in cancer research.
• Classification of pathway information into four categories.
• Categorization of models into three major types and seven subcategories based on pathway data utilization and application strategies.

Main Results:

• Pathway information application strategies are illustrated for each subcategory.
• Advantages and disadvantages of different pathway-informed models are summarized.
• Commonly used pathway databases and interpretability techniques are identified.

Conclusions:

• The review provides a structured overview of pathway-informed deep learning in cancer research.
• It offers insights into model design, strategy selection, and identifies current challenges.
• This work aims to guide the development of more effective pathway-informed deep learning models.