PathEmb: Random Walk Based Document Embedding for Global Pathway Similarity Search.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Pathway similarity search is crucial for understanding biological relationships.
• Current methods face computational inefficiencies due to the graph isomorphism problem.

Purpose of the Study:

• Introduce PathEmb, a novel algorithmic framework for efficient and accurate pathway similarity search.
• Address the computational limitations of existing pathway search methods.

Main Methods:

• PathEmb represents pathways as 'documents' and uses a second-order random walk strategy to explore patterns.
• Signaling paths from random walks form 'sentences,' creating a 'document' for each pathway.
• Pathways are mapped to a low-dimensional feature space, enabling topology-free search.

Main Results:

• PathEmb demonstrates superior computational efficiency and search accuracy compared to existing methods.
• The algorithm effectively handles pathways of arbitrary structure.
• Extensive evaluations on three pathway datasets validate PathEmb's performance.

Conclusions:

• PathEmb offers a significant advancement in pathway similarity search.
• The topology-free approach and improved efficiency make it a valuable tool for systems biology research.