Getting Started

Introduction

A high-throughput sequencing experiment typically identifies a large number of microbial features, also called taxa. High-dimensional statistics supplies estimators built for that regime. They assess interactions between taxa and reveal patterns that may be indicative of biological phenomena — shifts in community composition, associations with environmental factors.

The two plugins

QIIME2 ships a collection of plugins that preprocess raw sequences into feature tables ready for downstream analysis. For advanced statistical analysis of those tables, you often have no out-of-the-box option.

Fig. 1 High-dimensional statistics with QIIME2.

The pipeline above is implemented by two QIIME2 plugins: q2-gglasso for network learning and q2-classo for regression and classification. Together they close gaps in QIIME2, adding:

• Network inference through sparse inverse covariance estimation, which identifies microbial associations, for example which taxa co-occur within a community

• Sparse log-contrast models for classification and regression, which predict a measured covariate, such as disease status or an environmental factor, from the counts

• Compositionally-aware estimators that account for the constraints and the sparsity of microbiome counts

• Interactive visualizations with publication-ready network graphs and model performance plots

• QIIME2 integration using standard artifact formats and command-line interfaces

Scope

These chapters cover:

1. Install and set up the q2-gglasso and q2-classo plugins

2. Apply graphical lasso methods for network inference

3. Use log-contrast models for regression and classification

4. Interpret results and create publication-ready visualizations

5. Integrate these methods into reproducible QIIME2 workflows