|
Activity Number:
|
406
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #303941 |
|
Title:
|
Robust Regression for Detecting Copy Number Variants
|
|
Author(s):
|
Glen A. Satten*+ and Andrew Allen and Jennifer Mulle and Morna Ikeda and Stephen Warren
|
|
Companies:
|
CDC and Duke University and Emory University and Emory University and Emory University
|
|
Address:
|
MS K-23, Atlanta, GA, 30341,
|
|
Keywords:
|
Genetics ; Copy Number Variant ; CNV
|
|
Abstract:
|
The discovery of frequent deletions or repetitions (copy number variants or CNVs) in the human genome raises the question of how CNVs affect complex traits. Numerous algorithms for finding CNVs have been proposed. We propose a robust regression algorithm with a novel backward elimination for finding copy number variants (CNVs) using intensity data. User-specified parameters correspond to simple quantities like the smallest allowable CNV and the smallest change in median intensity (given the distance to the nearest feature) retained in the selection algorithm. Our algorithm can analyze data with 200,000 probes in 30 seconds on a laptop. Because our algorithm is fast, we develop permutation-based inference on CNVs by scoring each CNV and then comparing observed scores to scores obtained in null data. Null data are simulated using an ARMA model based on data with little or no signal.
|
