Multidimensional scaling improves distance-based clustering for microbiome data.

Area of Science:

• Microbiome research
• Computational biology
• Statistical genetics

Background:

• Patient subgrouping via microbial composition is vital for understanding health and disease.
• Distance-based clustering, like Partitioning Around Medoids (PAM), is computationally efficient but struggles with sparse microbial signal data.

Purpose of the Study:

• To enhance the performance of distance-based clustering methods for microbiome data.
• To address the limitations of existing methods in sparse signal scenarios.

Main Methods:

• A two-step procedure involving classical multidimensional scaling (MDS) for dimensionality reduction.
• Applying distance-based clustering on the low-dimensional data obtained from MDS.

Main Results:

• Multidimensional scaling (MDS) effectively denoises microbiome data.
• The proposed MDS-based procedure significantly improves clustering performance in sparse signal scenarios compared to direct distance-based clustering.
• The method's efficacy is validated through simulations and real-world microbiome data applications.

Conclusions:

• Classical multidimensional scaling (MDS) is a powerful preprocessing step for microbiome data clustering.
• The proposed method offers a robust solution for identifying microbial patterns related to human health and disease.