An iterative data mining approach for mining overlapping coexpression patterns in noisy gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Mining

Background:

• Traditional clustering algorithms often assume non-overlapping partitions, which is insufficient for gene expression data.
• Genes can coexpress with multiple groups due to complex biological interactions and responses to stimuli.
• Discovering overlapping clusters in noisy gene expression data presents a significant challenge for existing methods.

Purpose of the Study:

• To propose an iterative data mining approach for discovering overlapping clusters in gene expression data.
• To enhance the performance of existing clustering algorithms in bioinformatics.
• To effectively identify coexpressed genes that may belong to multiple functional groups.

Main Methods:

• An iterative two-phase approach combining initial non-overlapping clustering with iterative refinement.
• Phase 1: Initial partitioning of gene expression profiles using a standard clustering algorithm.
• Phase 2: Redetermining gene partition memberships using pattern discovery to allow for overlapping clusters.

Main Results:

• The proposed approach effectively discovers overlapping clusters in noisy gene expression data.
• Experimental results demonstrate improved performance compared to existing clustering algorithms.
• Validated on both artificial and real-world biological datasets.

Conclusions:

• The iterative data mining approach successfully addresses the limitations of non-overlapping clustering for gene expression analysis.
• This method offers a more biologically relevant way to identify coexpressed genes by accommodating overlapping functional relationships.
• The approach provides a valuable tool for advancing the understanding of gene regulation and biological pathways.