Activity Number:
|
76
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, July 31, 2016 : 4:00 PM to 5:50 PM
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
Abstract #319547
|
|
Title:
|
Detecting Association to Precision Networks via Conditional Multi-Type Graphical Models
|
Author(s):
|
Yanming Li* and Kevin He and Jian Kang and Hyokyoung (Grace) Hong and Ji Zhu and Yi Li
|
Companies:
|
University of Michigan and University of Michigan and University of Michigan and Michigan State University and University of Michigan and University of Michigan
|
Keywords:
|
conditional graphical model ;
precision network ;
high-dimensional data ;
cancer genomics
|
Abstract:
|
Understanding how genes regulate each other and how their regulations are associated with genomic markers with respect to individual patients can help uncover the mechanism of underlying biological or disease process at DNA level. Conditional graphical models are commonly used in simultaneously learning the gene regulatory network and recovering the association signals. Most of current conditional graphical models assume a homogeneous response network structure and only model the responses conditional means on the predictors. Also the current methods are not able to handle multi-type responses networks, therefore limit their applications in modern biomedical studies. We propose a multi-type conditional graphical model which allows heterogeneous patient level responses networks, mixture of types of responses and can accurately and effectively recover the associations of responses networks to high-dimensional biomarkers. The proposed method is computationally inexpensive and its finite sample properties are investigated both theoretically and empirically. We apply the method to the TCGA data to better understand the cancer-triggering biological pathways at molecular level.
|
Authors who are presenting talks have a * after their name.