Activity Number:
|
40
- Statistical Methods for Microbiome and Tumor Data
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, July 28, 2019 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
Abstract #307161
|
|
Title:
|
A Bayesian Framework for Uncovering Association Between Microbial Composition and Host Phenotypes
|
Author(s):
|
Subhajit Sengupta* and Riten Mitra and Robert Butler III and Abhishek Bhattacharjee and Pablo Gejman
|
Companies:
|
NorthShore University HealthSystem and University of Louisville and NorthShore University HealthSystem and University of Northern Colorado and NorthShore University HealthSystem
|
Keywords:
|
Bayesian analysis;
elastic-net;
microbiome;
phenotype;
sparse graphical model
|
Abstract:
|
Changes in microbiome community structure associate in many ways with host disease phenotype, presenting an enormous application in personalized medicine. Analyzing and interpreting microbiome data to gain meaningful biological insights remains a challenge due to the complex interactions within the community. Statistical models incorporating higher-order interactions are necessary to harmonize microbe-microbe interactions with host-microbe interactions for a meaningful phenotypic association. We develop a Bayesian sparse graphical model to associate microbial composition and phenotypes by disentangling interactions in microbiome. This neighborhood-based method uses elastic net algorithm to perform conditional sparse regression on each node. Incorporating multiple clinical groups coherent posterior inference is obtained on the differences among those groups. Thus, the key novelty of our approach is the use of interactions among operational taxonomic units (OTUs) to differentiate phenotype groups, rather than the differential distribution of OTUs alone. Using computationally efficient and scalable algorithm, we apply our method successfully to both synthetic and real microbiome data.
|
Authors who are presenting talks have a * after their name.