|
Activity Number:
|
428
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #307522 |
|
Title:
|
Variable Selection in the Multivariate Adaptive Regression Splines (MARS)-Logit Models To Detect Gene-Gene Interactions
|
|
Author(s):
|
Huiyi Lin*+ and Cruz Velasco and Jennifer J. Hu
|
|
Companies:
|
Universtiy of Alabama at Birmingham and Louisiana State University Health Sciences Center and Louisiana State University Health Sciences Center
|
|
Address:
|
WTI 153, 1530 3rd Ave S, Birmingham, AL, 35294,
|
|
Keywords:
|
variable selection ; gene-gene interaction ; MARS-logit ; multivariate adaptive regression splines ; simulation
|
|
Abstract:
|
Gene-gene interactions play critical roles in complex diseases. The MARS-logit models, which sequentially combine MARS and logistic models, can be applied for examining high-order interactions. A simulation study was carried out to assess the performance of the MARS-logit model using a number of variable selection approaches. Data on 20 single nucleotide polymorphisms (SNPs) were generated for 3000 subjects. Ten replicates were used to test 2- or 3-way interactions. Empirical guidelines for variable selection are provided. However, a software that combines both methods is needed for further evaluating the MARS-logit models. The MARS-logit models selected the correct SNPs, mode of inheritance and interaction type more often than MARS alone. Therefore, the MARS-logit hybrid model may be a better tool for exploring gene-gene interactions.
|
