Uncovering latent biological function associations through gene set embeddings.

Area of Science:

• Computational biology
• Systems biology
• Bioinformatics

Background:

• Biological network analysis uses graph theory to map interactions but struggles with incomplete data.
• Traditional methods often fail to integrate heterogeneous biological datasets effectively.
• The Molecular Signatures Database (MSigDB) provides attribute-driven knowledge crucial for enhancing network analysis.

Purpose of the Study:

• To extend conventional bipartite models by integrating attribute-driven knowledge.
• To explore unsupervised biological relationships and uncover gene-term associations.
• To develop a robust cross-species validation method for biological network analysis.

Main Methods:

• Integration of attribute-driven knowledge from MSigDB.
• Application of the node2vec algorithm for network embedding.
• Construction of a shared vector space for human and mouse data.

Main Results:

• Unsupervised discovery of biological relationships and gene associations.
• Successful cross-species validation of findings using embedded human and mouse data.
• Identification of both expected and novel biological insights.

Conclusions:

• The integrative framework offers a comprehensive perspective complementing traditional network analysis.
• This approach enhances the understanding of complex biological processes.
• The method paves the way for deeper insights into disease mechanisms.