Activity Number:
|
156
- Contributed Poster Presentations: Section on Statistics in Genomics and Genetics
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 8, 2022 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
Abstract #322417
|
|
Title:
|
A Bayesian Approach to Simultaneous Factorization and Prediction Using Multi-Omic Data
|
Author(s):
|
Sarah Samorodnitsky* and Chris Wendt and Eric F Lock
|
Companies:
|
University of Minnesota Division of Biostatistics and University of Minnesota Medical School and University of Minnesota
|
Keywords:
|
Omics;
Bayesian;
Multi-omics;
Lung Disease;
HIV;
COPD
|
Abstract:
|
Modern array-based technologies can magnify our view of a biological system by allowing us to collect multi-omic data. Integrative factorization methods are increasingly used to decompose such data into variation unique to each omics source and variation shared across sources. However, these methods often do not allow for prediction of a clinical outcome nor the quantification of uncertainty in estimating the latent variation structure. We address this gap by taking a Bayesian approach to factorization of multi-omic data that simultaneously incorporates outcome prediction. We use normal priors on the factorization components and show that the posterior mode of this model can be estimated by solving a structured nuclear norm penalized objective that also achieves rank selection (i.e., number of shared and unique components). This framework allows for concurrent imputation and full posterior inference for missing data, including “block-wise” missingness, and prediction of unobserved outcomes. We assess our model via simulation and apply it to multi-omic data from bronchoalveolar lavage to predict Chronic Obstructive Pulmonary Disease presence among patients who are HIV+.
|
Authors who are presenting talks have a * after their name.