Genetica

, Volume 140, Issue 10, pp 421–427

Power of a reproducing kernel-based method for testing the joint effect of a set of single-nucleotide polymorphisms

Authors

  • Hong He
    • Department of Epidemiology and Biostatistics, Norman J Arnold School of Public HealthUniversity of South Carolina
    • Department of Epidemiology and Biostatistics, Norman J Arnold School of Public HealthUniversity of South Carolina
  • Arnab Maity
    • Department of StatisticsNorth Carolina State University
  • Yubo Zou
    • Department of Epidemiology and Biostatistics, Norman J Arnold School of Public HealthUniversity of South Carolina
  • James Hussey
    • Department of Epidemiology and Biostatistics, Norman J Arnold School of Public HealthUniversity of South Carolina
  • Wilfried Karmaus
    • Department of Epidemiology and Biostatistics, Norman J Arnold School of Public HealthUniversity of South Carolina
Article

DOI: 10.1007/s10709-012-9690-5

Cite this article as:
He, H., Zhang, H., Maity, A. et al. Genetica (2012) 140: 421. doi:10.1007/s10709-012-9690-5

Abstract

This study explored a semi-parametric method built upon reproducing kernels for estimating and testing the joint effect of a set of single nucleotide polymorphisms (SNPs). The kernel adopted is the identity-by-state kernel that measures SNP similarity between subjects. In this article, through simulations we first assessed its statistical power under different situations. It was found that in addition to the effect of sample size, the testing power was impacted by the strength of association between SNPs and the outcome of interest, and by the SNP similarity among the subjects. A quadratic relationship between SNP similarity and testing power was identified, and this relationship was further affected by sample sizes. Next we applied the method to a SNP-lung function data set to estimate and test the joint effect of a set of SNPs on forced vital capacity, one type of lung function measure. The findings were then connected to the patterns observed in simulation studies and further explored via variable importance indices of each SNP inferred from a variable selection procedure.

Keywords

Reproducing kernels SNP Mixed linear models Testing power Variable selection

Copyright information

© Springer Science+Business Media Dordrecht 2012