Recommending pathway genes using a compendium of clustering solutions

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Gene expression data analysis commonly uses gene clustering, but often identifies only dominant coexpression groups.
• Recommender systems can predict new pathway members using known genes, but existing methods overlook gene co-clustering within experiments.

Purpose of the Study:

• To develop a novel collaborative filtering approach for identifying new pathway genes.
• To leverage gene clustering patterns across experiments for improved gene recommendation.

Main Methods:

• Genes are clustered within individual experiment series.
• Informative clusters containing user-query genes are identified.
• New genes are recommended if they co-cluster with known genes in a significant fraction of informative clusters.

Main Results:

• The proposed collaborative filtering method effectively recommends new genes for biological pathways.
• The new approach demonstrates comparable performance to established methods.
• Significant improvements in speed and scalability were achieved for large dataset searches.

Conclusions:

• Collaborative filtering based on gene clustering patterns offers an efficient and scalable solution for pathway analysis.
• This method enhances the discovery of novel pathway members from gene expression data.