A highly adaptive microbiome-based association test for survival traits

Hyunwook Koh, Alexandra E. Livanos, Martin J. Blaser, Huilin Li

Research output: Contribution to journalArticlepeer-review

11 Scopus citations


Background: There has been increasing interest in discovering microbial taxa that are associated with human health or disease, gathering momentum through the advances in next-generation sequencing technologies. Investigators have also increasingly employed prospective study designs to survey survival (i.e., time-to-event) outcomes, but current item-by-item statistical methods have limitations due to the unknown true association pattern. Here, we propose a new adaptive microbiome-based association test for survival outcomes, namely, optimal microbiome-based survival analysis (OMiSA). OMiSA approximates to the most powerful association test in two domains: 1) microbiome-based survival analysis using linear and non-linear bases of OTUs (MiSALN) which weighs rare, mid-abundant, and abundant OTUs, respectively, and 2) microbiome regression-based kernel association test for survival traits (MiRKAT-S) which incorporates different distance metrics (e.g., unique fraction (UniFrac) distance and Bray-Curtis dissimilarity), respectively. Results: We illustrate that OMiSA powerfully discovers microbial taxa whether their underlying associated lineages are rare or abundant and phylogenetically related or not. OMiSA is a semi-parametric method based on a variance-component score test and a re-sampling method; hence, it is free from any distributional assumption on the effect of microbial composition and advantageous to robustly control type I error rates. Our extensive simulations demonstrate the highly robust performance of OMiSA. We also present the use of OMiSA with real data applications. Conclusions: OMiSA is attractive in practice as the true association pattern is unpredictable in advance and, for survival outcomes, no adaptive microbiome-based association test is currently available.

Original languageEnglish (US)
Article number210
JournalBMC genomics
Issue number1
StatePublished - Mar 20 2018
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Biotechnology
  • Genetics


  • Community-level association test
  • High-dimensional compositional data analysis
  • Microbial group analysis
  • Microbiome-based association test
  • Microbiome-based survival analysis
  • Phylogenetic tree


Dive into the research topics of 'A highly adaptive microbiome-based association test for survival traits'. Together they form a unique fingerprint.

Cite this