RCM

RCM combines log-linear analysis of contingency tables, flexible response function estimation, and empirical Bayes dispersion estimation to provide unconstrained and constrained ordination for explorative visualization of microbiome read count datasets.

Key Features:

• Unconstrained ordination: Performs dimension reduction for explorative visualization of microbiome datasets without predefined constraints.
• Constrained ordination: Performs ordination with correction for observed technical confounders within the same framework.
• Flexible response function modelling: Models relationships between bacterial species and ordination axes using flexible response functions.
• Empirical Bayes dispersion estimation: Estimates dispersion to account for overdispersion and differences in dispersion between taxa.
• Sequencing depth adjustment: Copes with varying sequencing depths across samples.
• Diagnostic plots: Provides plots to detect lack of fit and verify explicit assumptions.

Scientific Applications:

• Microbiome data exploration: Yields meaningful biological patterns from microbiome read count datasets through dimension reduction.
• Bacterial species role elucidation: Identifies and visualizes the role of bacterial species in microbiome datasets, unlike methods that focus only on sample ordination.
• Technical confounder correction: Corrects for observed technical confounders that adversely affect other ordination methods.
• Method comparison and validation: Enables identification of flaws in existing distance-based ordination methods through explicit, verifiable assumptions.

Methodology:

RCM combines log-linear models with a dispersion estimation algorithm and flexible response function modelling into a unified framework for ordination, implemented as an R package using the negative binomial distribution.