banocc

banocc estimates sparse precision matrices from compositional data using a Bayesian framework to infer component correlations and quantify posterior uncertainty for ecological and microbiome studies.

Key Features:

• Bayesian framework: Uses Bayesian inference to estimate the precision matrix of compositional data while quantifying posterior uncertainty.
• LASSO prior for sparsity: Employs a LASSO prior to identify and estimate sparse structures in high-dimensional precision matrices.
• MCMC sampling: Leverages Markov Chain Monte Carlo sampling to generate posterior distributions and quantify uncertainty of functions derived from the precision matrix, including correlation matrices.
• Transformation approximation: Applies a first-order Taylor expansion to approximate the transformation from unobserved counts to compositions for correlation inference from normalized proportions.
• Posterior inference and validation: Provides posterior-based inference and was validated on simulated datasets showing accurate network recovery with low type I and II error rates and competitive performance versus existing methods.

Scientific Applications:

• Ecological and microbiome network analysis: Infers component interactions and correlations in ecological research and microbiome studies using compositional sequencing data.
• Human Microbiome Project analysis: Applied to a Human Microbiome Project microbial ecology dataset, reproducing established findings and identifying competitive roles of Proteobacteria across habitats.

Methodology:

Approximates count-to-composition transformations using a first-order Taylor expansion; employs Bayesian inference with MCMC sampling to estimate a sparse precision matrix and quantify uncertainty; implements the Bayesian model in rstan.