Random coefficient regression models in QTL interval mapping

ALEJANDRA T. ARROYO; MÓNICA G. BALZARINI

Dublin

Congreso; XXIV International Biometric Conference; 2008

Quantitative trait loci (QTL) analysis has changed the conventional view of polygenic inheritance. Several strategies have been proposed to explore associations between phenotypic values and molecular marker genotypes leading to the identification of loci producing variation in continuous traits. Composite interval mapping (CIM) is a common method of QTL analysis in plant genetics. In this method, the phenotypic response is regressed on the expected value of a random variable indicating the most likely QTL genotype given the genotype of two molecular markers flanking a genome interval where the QTL is supposed to be placed. Markers outside that interval are used as covariables (cofactors) in the model to control genetic background.   The distribution of model errors is approximated as a mixture of normal distributions depending on the number of possible genotypes at the putative QTL in the mapping population. ML estimation of full and reduced (QTL effect equals zero at the position under analysis) models provides statistical support to detect QTL effects and QTL positions on the genome. CIM is usually run with previous variable (cofactor) selection to avoid redundancy. However, if the experimental plant population has several replicates of each plant genotype, a genotype effect could be used in addition to cofactors to control the genetic background not accounted for the markers already in the analysis.