Response projected clustering for direct association with physiological and clinical response data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray gene expression data analysis often lacks direct integration of physiological and clinical metadata.
• Current clustering methods cannot effectively incorporate quantitative metadata into gene expression heatmaps.
• There is a need to intuitively discover gene groups relevant to disease phenotypes using clinical response data.

Purpose of the Study:

• To introduce a novel clustering analysis approach, Response Projected Clustering (RPC), for integrating gene expression and clinical metadata.
• To enable intuitive discovery of gene networks associated with disease phenotypes by summarizing clinical response data.
• To enhance the analysis of high-dimensional microarray data with quantitative metadata.

Main Methods:

• Developed Response Projected Clustering (RPC), a method using high-dimensional geometrical projection of response data into gene expression space.
• Clustered projected gene vectors with the projected response vector based on association degrees.
• Employed bootstrap-counting for RPC analysis to evaluate statistical tightness of gene clusters.

Main Results:

• RPC was applied to NCI-60 cancer cell line data and macrophage differentiation in atherogenesis studies.
• Identified known and novel gene factors and potential pathway associations relevant to drug chemosensitivity and atherogenesis.
• Demonstrated RPC's ability to effectively summarize individual genes' association with metadata and expression patterns.

Conclusions:

• Response Projected Clustering (RPC) effectively discovers gene networks associated with clinical metadata.
• RPC enhances the utility of clustering analysis for high-dimensional microarray gene expression data.
• The method facilitates intuitive discovery of gene-phenotype relationships.