Challenges for case-control studies with microbiome data

Abstract

Purpose: In case-control studies of the human microbiome, the goal is to evaluate whether cases differ from controls in the microbiome composition of a particular body habitat and which taxa are responsible for the differences. These studies leverage sequencing technology and spectroscopy that provide new measurements of the microbiome.

Methods: Three challenges in conducting reproducible microbiome research using a case-control design are compensating for differences in observed and actual microbial community composition, detecting "rare" taxa in microbial communities, and choosing properly powered analysis methods. The significance of each challenge, evaluation of commonly held views, analysis of unanswered questions, and suggestions of strategies for solutions are discussed.

Results: Understanding the effects of these choices on case-control analyses has been underappreciated, with an implicit assumption that further advances in technology will address all the current shortcomings.

Conclusions: It is recommended that research on the human microbiome include positive and negative control experiments to provide insight into bias, contamination, and technical variation. Research protocols such as these may afford a better opportunity to make quantitative and qualitative adjustments to data, thereby reducing the risk of falsely positive results, increasing power to discover true disease determinants, and enhancing interpretation across studies.

Keywords: Bias; Control experiments; Microbiome; Normalization; Rare taxa.