Reordering hierarchical tree based on bilateral symmetric distance.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Mining

Background:

• Hierarchical clustering (HC) is widely used in microarray data analysis to group samples or genes based on expression profiles.
• Traditional HC methods create tree structures but obscure individual object relationships, leading to interpretation challenges.
• The loss of object-level relationships hinders the understanding of associations within complex datasets.

Purpose of the Study:

• To propose a novel ordering method, HC-SYM, to enhance the interpretability of hierarchical clustering trees.
• To address the limitation of lost individual object relationships in standard HC visualizations.
• To facilitate more intuitive exploration of high-dimensional data.

Main Methods:

• Developed HC-SYM, an ordering algorithm that minimizes the bilateral symmetric distance between adjacent clusters.
• The method aims to position similar objects at the boundaries of clusters for clearer visualization.
• Evaluated HC-SYM performance using both supervised and unsupervised analytical approaches.

Main Results:

• HC-SYM demonstrated favorable performance when compared to existing ordering methods.
• The proposed method effectively improves the visualization of relationships between individual objects within clusters.
• Evaluation confirmed the efficacy of HC-SYM in both supervised and unsupervised learning contexts.

Conclusions:

• HC-SYM offers an intuitive understanding of object relationships within hierarchical clustering structures.
• The flexibility of HC-SYM makes it valuable for exploratory analysis of microarray and other high-dimensional data.
• This method enhances the utility of hierarchical clustering for biological and data science research.