T-BAPS

T-BAPS applies a Bayesian latent class mixture model with MCMC-based clustering to terminal-restriction fragment length polymorphism (T-RFLP) data to characterize and compare microbial community structure.

Key Features:

• Bayesian latent class mixture model: Implements a latent class mixture model that explicitly addresses sparseness, missing values, and structurally zero-valued observations in T-RFLP datasets.
• Markov Chain Monte Carlo (MCMC) inference: Uses MCMC-based Bayesian model selection and inference to estimate cluster assignments and model parameters.
• Benchmarking against conventional methods: Evaluated performance relative to k-means, hierarchical clustering, and Autoclass, with simulations showing comparable or superior results.
• Application to simulated and real datasets: Applied to both simulated and empirical T-RFLP datasets to assess applicability across research scenarios.

Scientific Applications:

• Microbial community characterization: Characterizes microbial community composition from T-RFLP molecular fingerprints.
• Comparative community analysis: Compares microbial populations across samples or environments using clustering results.
• Diversity and dynamics inference: Facilitates inference of microbial diversity, community structure, and temporal or spatial dynamics.
• Environmental, biotechnological, and public health studies: Supports analyses in environmental microbiology, biotechnology, and public health contexts.

Methodology:

Implements a Bayesian latent class mixture model that accounts for sparseness, missing values, and structurally zero observations, with model selection and inference performed via Markov Chain Monte Carlo (MCMC); performance was compared to k-means, hierarchical clustering, and Autoclass on simulated and real T-RFLP datasets.