Nonparametric covariance estimation in functional mapping of complex dynamic traits

摘要: One of the fundamental objectives in agricultural, biological and biomedical research is identification genes that control developmental pattern complex traits, their responses to environment, way these interact a coordinated manner determine final expression trait. More recently, new statistical framework, called functional mapping, has been developed identify map quantitative trait loci (QTLs) trajectories by integrating biologically meaningful mathematical models progression into mixture model for unknown QTL genotypes. Functional mapping emerged be powerful tool QTLs controlling responsiveness (reaction norm) environmental signals. From perspective, designed study genetic regulation network variation dynamic traits virtually joint mean-covariance likelihood model. Appropriate choices mean covariance structures are critical importance inference about locations actions/interactions. While battery have proposed vector modeling, analysis structure mostly limited parametric like autoregressive one (AR(1)) or structured antedependence (SAD) In reaction norms respond two signals, model, expressed as Kronecker product AR(1) structures, test differences different environments. For practical longitudinal data sets, modeling may too simple capture covariance. There pressing need develop robust approach any possible covariance, ultimately broadening use mapping. Our proposes nonparametric estimator locus. We adopt Huang et al.'s (2006) invoking modified Cholesky decomposition converting problem sequence regressions responses. A regularized positive-definite obtained using normal penalized with an L2 penalty. This embedded within framework reparameterized version derivative log-likelihood. extend idea interaction effects between signals non-separable way. The extended allows several questions. Is there pleiotropic regulates genotypic signals? What difference timing duration environment-specific responsiveness? How environment-dependent regulated development-related QTL? performed various simulation studies reveal properties demonstrate advantages estimator. By analyzing real examples studies, we illustrated utilization usefulness methodology. methods will provide useful genome-wide scanning existence, distribution interactions underlying important agriculture, biology health sciences.