Density of points clustering, application to transcriptomic data analysis.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-throughput technologies generate vast datasets requiring robust exploratory data analysis.
• Clustering algorithms are essential for grouping similar data points but often need manual parameter tuning, such as specifying the number of clusters.
• Automated determination of the optimal number of clusters remains a challenge in data mining and bioinformatics.

Purpose of the Study:

• To develop and validate a novel, automated stopping rule for determining the optimal number of clusters.
• To address the limitation of manual parameter setting in clustering algorithms.
• To apply the method to analyze complex biological data, specifically transcriptomic data.

Main Methods:

• A novel stopping rule based on comparing intra-cluster and inter-cluster point densities was developed.
• The proposed method was evaluated using synthetic datasets and real-world transcriptomic data.
• Performance was benchmarked against two existing clustering evaluation methods.

Main Results:

• The novel stopping rule effectively identified the optimal number of clusters across various datasets.
• Comparison with existing methods demonstrated competitive or superior performance.
• Application to promyelocytic cell data revealed insights into retinoic acid signaling pathways.

Conclusions:

• The developed density-based stopping rule offers an effective and automated approach to determining the optimal number of clusters.
• This method enhances the analysis of high-throughput biological data, facilitating the discovery of complex biological mechanisms.
• The approach has significant implications for gene expression analysis and understanding cellular responses to treatments.