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 #321248
|
View Presentation
|
Title:
|
Longitudinal Gaussian Graphical Models for Autism Risk Gene Detection
|
Author(s):
|
Kevin Lin*
|
Companies:
|
Carnegie Mellon University
|
Keywords:
|
Graphical Model ;
Autism ;
High-Dimensional Inference ;
Microarray ;
Meta-Analysis
|
Abstract:
|
Previous studies have shown the importance of using gene-pair interactions when detecting genes influencing autism spectrum disorder. The conditional-independence graphical model has been an natural tool to represent this gene network. This graph has been previously estimated from microarray data originating from brain tissue belonging a specific spatio-temporal window. This window was chosen to match a key developmental time and region of the brain associated with autism, but the corresponding number of samples is extremely limited. We improve the graphical model estimation in two ways. 1) We test for relevant microarray samples originating from other spatio-temporal windows and aggregate these selected samples using longitudinal transformation. This graphical model meta-analysis increases the sample size while preserving homogeneity among the samples. 2) We prune the non-significant edges of the network via high-dimensional hypothesis testing. This removes spurious edges. We demonstrate the success of our method by improvements in the number of genes also found in an independent study and stability of our results with respect to tuning parameters.
|
Authors who are presenting talks have a * after their name.