Differential co-expression framework to quantify goodness of biclusters and compare biclustering algorithms

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Biclustering is crucial for understanding biological mechanisms in gene expression data.
• Existing biclustering algorithms lack standardized performance comparison and unified ranking methods on real datasets.

Purpose of the Study:

• To develop a differential co-expression framework for objective bicluster quality assessment.
• To create a scoring function for quantifying bicluster goodness based on differential co-expression patterns.
• To evaluate and compare the performance of established biclustering algorithms using real-world gene expression data.

Main Methods:

• Proposed a differential co-expression framework and scoring function to evaluate bicluster quality.
• Genes within a bicluster are expected to be co-expressed in specific conditions and not in others.
• Applied scoring functions to analyze four biclustering algorithms on six diverse real datasets, creating a unified ranking.

Main Results:

• The differential co-expression framework provides objective assessment of bicluster quality.
• A scoring function was developed to stratify biclusters into three co-expression types.
• Performance and behavior of four biclustering algorithms were analyzed and compared through unified ranking.

Conclusions:

• The differential co-expression framework offers a quantitative and objective method for assessing biclusters.
• It enables effective comparison of biclustering algorithm performance in identifying co-expressed gene sets.
• Facilitates meta-biclustering by combining outputs from different algorithms into a single, ranked list.