pyGCluster, a novel hierarchical clustering approach.

Area of Science:

• Computational biology
• Bioinformatics

Background:

• Clustering algorithms are essential for analyzing large biological datasets.
• Assessing the reproducibility of clustering results is crucial for reliable biological insights.

Purpose of the Study:

• To introduce pyGCluster, a Python-based clustering algorithm.
• To evaluate the reproducibility of agglomerative hierarchical clustering.
• To develop a method for meta-clustering reproducible clusters into communities.

Main Methods:

• Implemented a clustering algorithm with a focus on noise injection for validation.
• Assessed reproducibility across various distance-linkage combinations in hierarchical clustering.
• Developed a meta-clustering approach to form communities from reproducible clusters.
• Created graphical representations including node and expression maps.

Main Results:

• Demonstrated the utility of noise injection for cluster validation.
• Identified reproducible clusters within large datasets.
• Successfully meta-clustered reproducible clusters into meaningful communities.
• Provided visual tools for result interpretation.

Conclusions:

• pyGCluster offers a robust approach to cluster validation and analysis.
• The algorithm enhances the reliability of findings from hierarchical clustering.
• Meta-clustering reproducible clusters facilitates the discovery of higher-order biological structures.