Tidi_MIBI: a Tidy Pipeline for Microbiome Analysis and Visualization in R

• Activity Number: 531 - SPEED: Statistical Computing: Methods, Implementation, and Application, Part 2
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 11:35 AM to 12:20 PM
• Sponsor: Section for Statistical Programmers and Analysts
• Author(s): Charlie Carpenter*
• Companies: University of Colorado-Biostatistics
• Keywords: Microbiome; R; tidyverse; pipeline; visualization; negative binomial modeling

Abstract

We have developed an R package called tidi_MIBI to streamline microbiome management and analysis within the “tidy” framework developed by Hadley Whickham. The package provides a single, built in function that formats OTU tables from multiple taxonomic levels, calculates taxa prevalence, relative abundance, and a CLR transformation, and finally merges in clinical data. It also includes functions for basic ecological calculations and plots. A main feature of tidy_MIBI is negative binomial (NB) regression models for taxa relative abundance. With one line of code users can fit NB models to each taxa within a taxonomic rank with options to include random effects. The function returns user-friendly model summaries including the estimate rate ratios, confidence intervals, and FDR corrected p-values. A unique inclusion for tidi_MIBI is the inclusion of stacked bar charts of estimated relative abundance based on NB models, a novel and powerful tool for visualizing the results of dozens of models simultaneously. We demonstrate the usability of tidi_MIBI through application to an existing study on maternal obesity and gestational diabetes and the infant microbiome.