Finding Quantitative Trait Loci Genes in Experimental Crosses with Targeted Maximum Likelihood Estimation in a Semiparametric Model

• Activity Number: 30
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Hui Wang*+
• Companies: University of California, Berkeley
• Keywords: Targeted maximum likelihood ; Semiparametric model ; QTL mapping ; Experimental crosses

Abstract

Classic quantitative trait loci (QTL) mapping methods assume parametric models, and parameter estimation relies on maximum likelihood approach. We propose a semiparametric model with targeted MLE (tMLE) to estimate and test QTL parameters. Semiparametric models require few assumptions; however, MLE estimates are often biased in this class of models. To solve this problem, tMLE is used for parameter estimation. It is an extension of MLE targeted towards the parameter of interest. There are two stages involved in our tMLE procedure. The 1st stage is the estimation of the overall density of the phenotype given the QTL genotype. The 2nd stage adjusts for the first-stage estimator using nearby markers confounded with the QTL. Estimators of tMLE are often more accurate than those from MLE, and enjoy favorable statistical properties such as double robustness and local efficiency.