CLAG: an unsupervised non hierarchical clustering algorithm handling biological data.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Clustering biological data presents challenges, including data points that resist single clustering or belong to multiple clusters.
• Traditional methods like hierarchical agglomerative clustering and supervised classification are often unsuitable for complex or poorly understood datasets.

Purpose of the Study:

• To introduce CLAG (CLusters AGgregation), an unsupervised, non-hierarchical clustering algorithm.
• To develop a method capable of clustering diverse biological data types while ensuring consistent and reliable results.

Main Methods:

• CLAG is an unsupervised, non-hierarchical clustering algorithm.
• It processes correlation matrices (protein families, gene expression, miRNA), multidimensional species data, and binary matrices.
• The algorithm does not require all data points to form clusters and guarantees convergence to the same result upon repeated runs.

Main Results:

• CLAG successfully clusters various biological datasets, including protein families, cancer-related gene expression and miRNA data, species character data, and binary matrices.
• The algorithm provides a clustered matrix and numerical values indicating cluster strength.
• CLAG demonstrates simplicity and speed, making it suitable for reasonably large datasets.

Conclusions:

• CLAG is effective for exploring cluster structures and identifying underlying graphs in biological datasets.
• It surpasses established methods like hierarchical agglomerative clustering, k-means, fuzzy c-means, model-based clustering, and affinity propagation.
• CLAG avoids convergence issues common in some other clustering algorithms.