ARBic

ARBic identifies significant biclusters in large-scale gene expression datasets to uncover functionally correlated genes and reveal underlying biological processes and pathways.

Key Features:

• Comprehensive bicluster identification: Detects meaningful biclusters of any shape—broad or narrow—even when values within biclusters share similar distributions with background data.
• Integrated column- and row-based strategies: Combines a column-based strategy inspired by ReBic for narrow biclusters with a row-based strategy developed for ARBic to preferentially identify broader biclusters.
• Graph-based row strategy: The row-based approach repeatedly finds the longest path in a directed graph to detect broader biclusters.
• Evaluation performance: On simulated datasets and comparisons with seven other biclustering algorithms, achieved 29% higher recovery, relevance, and F1-scores than the second-best algorithm and outperformed others on real datasets.
• Robustness: Demonstrates robustness to noise, diverse bicluster shapes, and different types of datasets.

Scientific Applications:

• Gene co-expression discovery: Reveals functionally correlated genes from gene expression matrices.
• Pathway and process elucidation: Aids understanding of biological processes and molecular pathways by grouping co-regulated genes.
• Gene network inference: Facilitates discovery of gene networks through identification of co-expressed gene modules.
• Biomarker identification: Supports identification of potential biomarkers for diseases by isolating condition-specific biclusters.
• Cross-context applicability: Applies across diverse research contexts and dataset types, from basic genomics studies to applied biomedical research.

Methodology:

Integrates a column-based strategy inspired by ReBic and a row-based strategy that repeatedly finds the longest path in a directed graph to identify biclusters of varying shapes.