|
Activity Number:
|
518
- Statistical Methods for Complex Interactions and Genetic and Environmental Epidemiology
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #304758
|
Presentation
|
|
Title:
|
Autoregressive Zero Inflated Mixed-Effect Model on Time Series Microbiome Data
|
|
Author(s):
|
Linchen He* and Huilin Li
|
|
Companies:
|
New York University and NYU School of Medicine
|
|
Keywords:
|
Time series data;
interaction network;
microbial stability;
high-dimentional;
16S rDNA;
absolute abundance
|
|
Abstract:
|
High-throughput sequencing technology has been used to profile and characterize the taxonomic composition of microbial communities. More and more time-series data from 16S rDNA amplicon sequencing become available in recent microbiome studies. To investigate the microbial interactions and stability, we proposed an autoregressive zero-inflated mixed-effect model (ARZIMM) to model the excess zero abundance and the non-zero abundances separately and to use a random effect model to borrow strength across subjects. Based on the microbial interaction estimates from ARZIMM, we further investigated stability properties of microbial community by quantifying the rate of the transition distribution approaching the stationary distribution as well as the immediate response of a system following a perturbation. In the extensive simulation studies and a real data analysis, ARZIMM outperformed the competing methods and exhibited its feasibility for examining microbial interactions and stability based on microbial time-series data.
|
Authors who are presenting talks have a * after their name.
