Exploratory Analysis of the Gene Expression Matrix Based on Dual Conditional Dimensionality Reduction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression data analysis aims to find relationships between genes and biological conditions.
• Algorithmic results require contextual analysis due to confounding factors like tissue differentiation.
• Visual analytics is crucial for exploratory analysis of gene expression matrices (GEM) in biomedical research.

Purpose of the Study:

• To present a visual analytics approach for discovering connections between genes and samples in gene expression data.
• To enable contextual framing of analysis within the user's domain knowledge.
• To facilitate the identification of meaningful relationships in complex biological datasets.

Main Methods:

• Developed a visual analytics approach linking a reordered GEM heatmap with dual 2D projections.
• Enabled recomputation of projections conditioned by user-selected subsets of genes and/or samples.
• Utilized interactive exploration of gene expression data.

Main Results:

• Demonstrated the approach's capability to uncover relevant biological knowledge.
• Successfully applied the method to TCGA data involving two cancer types and normal tissue.
• Facilitated the discovery of gene-sample relationships.

Conclusions:

• The visual analytics approach effectively aids in exploring gene expression data.
• The method enhances the discovery of meaningful connections between genes and biological conditions.
• This approach is valuable for biomedical research, particularly in cancer studies.