SSMoRe: A Method for Sample Size Calculation in Microbiome Epidemiology

• Activity Number: 29 - Advances in Methods for Microbiome and Omics Data
• Type: Contributed
• Date/Time: Sunday, August 7, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Meghan Short* and Paulo Manrique and Lauren McIver and Jeremy E. Wilkinson and Eric A. Franzosa and Curtis Huttenhower
• Companies: Harvard T. H. Chan School of Public Health and Harvard T. H. Chan School of Public Health and Harvard T. H. Chan School of Public Health and Harvard T. H. Chan School of Public Health and Harvard T. H. Chan School of Public Health and Harvard T.H. Chan School of Public Health
• Keywords: microbiome; epidemiology; compositional data; power; sample size

Abstract

Tests of differential feature abundance are a cornerstone of molecular epidemiology, and they have proven particularly challenging to develop accurately for the human microbiome. Because of the complexity of tests needed for microbial measurements as compared to other ‘omics profiles, corresponding reliable methods for power and sample size estimation are lacking. We have developed both empirical and analytical means to address these gaps in a new method, Sample Sizes for Microbiome Research (SSMoRe). These build on a simulation framework developed for microbial community structure using zero-inflated log-linear models fit to ecologies of interest to subsequently generate realistic synthetic feature tables. We estimated power for various effects and sample sizes via simulations. We identified strong relationships between power and feature prevalence unaccounted for in standard formulae. Use of an “effective” sample size adjusted for feature prevalence improved power formula accuracy (i.e. similarity to simulated power). We plan to continue validation of these methods using resampling techniques in studies of the human microbiome spanning a range of health and disease phenotypes.