New Methodology And Its Application In Nested Association Mapping

Classical statistical approaches for quantitative trait locus (QTL) detection were established on bi-parental segregation population. Among these methods, there are two shortcomings. One is that the number of alleles at one locus for bi-parental segregation population is at most two. If the two parents share one same allele, the locus can not be detected even though its effect is large enough.Especially, the results of QTL mapping vary from population to population. The other is that genetic population is different from breeding population. This results in a limited value of QTL mapping in crop breeding. To overcome these shortcomings, the scientists at Cornell University developed a nested association mapping approach.Nested association mapping was created on the basis of a reference design where one inbred line was crossed to twenty-five diverse inbred lines and two hundred recombinant inbred lines were derived from each cross. This design can integrate the advantages of linkage and association analyses, overcome the shortcomings of traditional linkage analysis, and increase the power of QTL detection. This idea has been widely known. Based on previous studies, genome-wide muti-QTL and epistatic QTL mapping approaches for nested association mapping have been proposed in this study. The model parameters were estimated by empirical Bayes and expectation-maximization lasso algorithms. A series of simulation data along with real data in maize nested assocaition mapping was used to validate the above approaches. The main results were as follows.A series of Monte Carlo simulation experiments were used to validate the proposed approaches implemented by empirical Bayes algorithm. Results showed that false positive rate is only0.3%; the estimates for QTL variance and position were unbiased; and the power for QTL detection was high, for example, almost100%for the QTL with the proportion of phenotypic variance explained of5%, and more than80%for the QTL with the proportion of0.75%. The proposed approaches in this study were used to analyze real datasets of flowering time traits in maize nested association mapping. As a result,41main-effect QTL and23epistatic QTL for days to anthesis;34main-effect QTL and19epistatic QTL for days to silk; and43main-effect QTL and9epistatic QTL for anthesis-silking interval were detected using expectation-maximization lasso algorithm. The total proportions of phenotypic variances explained (PVE) by all the QTL for the three traits were72.87,70.35and68.97%, respectively; the total PVE by main-effect QTLwere28.16,23.66and58.61%, respectively, and the average PVE were0.69,0.70and1.36%, respectively; the total PVE by epistatic QTL were44.71,46.69and10.36%, respectively, and the average PVE were1.94,2.46and1.15%, respectively. This indicates that the relative PVE by main-effect or epistatic QTL varies from trait to trait.