BiCluE

BiCluE performs bicluster editing, including the Bi-Force heuristic, to identify local co-clustering patterns in high-dimensional biological data using exact and heuristic algorithms for both weighted and unweighted problems.

Key Features:

• Biclustering approach: Employs simultaneous clustering (co-clustering) to uncover local patterns in high-dimensional biological data such as gene expression datasets.
• Weighted and unweighted bicluster editing: Supports both weighted and unweighted formulations of the bicluster editing problem.
• Exact and heuristic algorithms: Provides exact algorithms alongside heuristic methods for solving bicluster editing instances.
• Bi-Force heuristic: Introduces the Bi-Force heuristic grounded in the weighted bicluster editing model to enhance solution capability for complex bicluster editing problems.
• Pairwise similarity support: Leverages any type of pairwise similarities between entities as input to the biclustering process.
• Comparative benchmarking: Was evaluated against FABIA, QUBIC, Cheng and Church, Plaid, BiMax, Spectral, xMOTIFs, and ISA using synthetic datasets and nine Gene Expression Omnibus (GEO) gene expression datasets following the protocol of Eren et al. (2013).
• Biological relevance assessment: Assesses resulting clusters via Gene Ontology enrichment analysis to evaluate biological significance.
• Implementation: Implemented in Java and integrated into the BiCluE software package.

Scientific Applications:

• Gene expression analysis: Identification of local co-expression patterns and biclusters in GEO gene expression datasets.
• Functional module discovery: Detection of biologically relevant modules assessed through Gene Ontology enrichment analysis.
• Biomedical high-dimensional data analysis: Analysis of local interactions within large-scale biomedical datasets using weighted or unweighted bicluster editing formulations.

Methodology:

Uses biclustering (simultaneous/co-clustering) with weighted bicluster editing and both exact algorithms and the Bi-Force heuristic, accepts arbitrary pairwise similarities, was benchmarked against eight established biclustering tools on synthetic and nine GEO gene expression datasets following Eren et al. (2013), and evaluates clusters via Gene Ontology enrichment analysis; implemented in Java.