Comparison of beta diversity measures in clustering the high-dimensional microbial data

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Disease heterogeneity complicates medical research, necessitating patient subgroup classification.
• Gut microbiome composition is increasingly recognized for its role in disease pathogenesis and patient prognosis.
• Clustering patients by microbial profiles requires appropriate beta diversity measures, but selecting the optimal measure is challenging.

Purpose of the Study:

• To evaluate the performance of various beta diversity measures for clustering microbial compositional data.
• To identify robust beta diversity measures that accurately reflect compositional changes, especially in low-abundance microbial taxa.
• To provide guidance on selecting appropriate measures for microbiome-based patient stratification.

Main Methods:

• Simulation experiments were designed to mimic high-dimensional microbial compositional data from 16S rRNA sequencing.
• Performance evaluation of multiple beta diversity measures, including Kullback-Leibler divergence-based (Jensen-Shannon) and hypersphere-based (Bhattacharyya, Hellinger) metrics.
• Validation of simulation findings using two real-world microbiome datasets.

Main Results:

• Kullback-Leibler divergence-based (Jensen-Shannon) and hypersphere-based (Bhattacharyya, Hellinger) beta diversity measures demonstrated superior performance.
• These selected measures efficiently captured compositional shifts, particularly in low-abundance microbial elements.
• The measures exhibited stable performance across simulation and real-world datasets, confirming their reliability.

Conclusions:

• Jensen-Shannon divergence and Bhattacharyya/Hellinger distances are recommended for clustering microbial compositional data.
• These measures offer robust and stable patient stratification by effectively handling the complexities of microbiome data.
• Accurate microbiome-based patient clustering can significantly advance the understanding and treatment of complex diseases.