fast.adonis

fast.adonis performs non-parametric multivariate analysis on dissimilarity matrices to identify variables associated with microbial community differences and quantify their contributions while accommodating sampling weights for large-scale microbiome studies.

Key Features:

• Non-parametric multivariate analysis: Performs multivariate analysis on dissimilarity matrices to identify explanatory variables and quantify their contributions.
• Compatibility with adonis/adonis2 (vegan): Produces results consistent with adonis and adonis2 from the vegan R package.
• Large-scale dataset support: Optimized for very large datasets (n ≥ 5000) where traditional methods become computationally intensive.
• Algebraic integration of sampling weights: Incorporates sampling weights algebraically to emulate the source population without replicating samples.
• Reduced permutation/bootstrap burden: Lowers computational cost associated with sequential multivariate analysis using permutations or bootstrapping.
• Lower memory requirements: Circumvents extensive memory demands typical of large-scale multivariate analyses.
• Complex sampling designs: Supports analyses that require adjustments for sampling weights to obtain unbiased estimates.
• R implementation: Implemented as an R function for integration into R-based analysis workflows.

Scientific Applications:

• Microbiome research: Applied to large microbiome datasets to assess relationships between microbial communities and explanatory variables.
• Variable contribution quantification: Used to quantify the contribution of covariates to dissimilarity in community composition.
• Complex survey or sampled studies: Supports studies with complex sampling designs that require weighted analyses for unbiased inference.
• Large-scale multivariate studies beyond microbiomes: Applicable to other domains requiring efficient non-parametric multivariate analysis on large datasets.

Methodology:

Implemented as an R function that algebraically integrates sampling weights to emulate the source population without sample replication and yields results consistent with adonis/adonis2 (vegan) while reducing the memory and computational burden of permutation- or bootstrap-based multivariate analyses.