Bipartite tight spectral clustering (BiTSC) algorithm for identifying conserved gene co-clusters in two species

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Gene clustering aids in predicting gene functions within a species.
• Predicting gene function across species using conserved genes remains a challenge.
• A novel computational algorithm is needed to identify co-clusters of evolutionarily conserved genes.

Purpose of the Study:

• To develop a new computational algorithm for identifying gene co-clusters in two species.
• To leverage gene orthology and expression data for cross-species gene function prediction.
• To create a flexible and robust algorithm for identifying conserved gene co-clusters.

Main Methods:

• Developed the bipartite tight spectral clustering (BiTSC) algorithm.
• Encoded gene orthology as a bipartite network and gene expression as node covariates.
• Integrated kernel enhancement, bipartite spectral clustering, consensus clustering, tight clustering, and hierarchical clustering.

Main Results:

• BiTSC identifies informative gene co-clusters without forcing all genes into clusters.
• The algorithm is flexible, robust, and makes no distributional assumptions.
• Demonstrated accuracy and robustness through simulations and identified conserved co-clusters in Drosophila melanogaster and Caenorhabditis elegans.

Conclusions:

• BiTSC accurately and robustly identifies conserved gene co-clusters across species.
• The algorithm enhances cross-species gene function prediction by utilizing evolutionary conservation.
• BiTSC offers a generalizable approach for co-cluster identification in bipartite networks.