A mixture model approach to the mapping of QTL controlling endosperm traits with bulked samples

Abstract

Endosperm traits are of triploid inheritance and have become a focus of breeding effort for their close relations with the grain quality. Current methods for mapping quantitative trait loci (QTL) underlying endosperm traits are restricted to the use of the phenotypes of single grain samples as input data set, which are often not available in practice due to the small size of the cereal seeds. This paper proposed a statistical model for one specially tailored mapping strategy, where the marker genotypes are obtained from the maternal plants in the segregation population and the phenotypic responses are replaced by the trait means of composite endosperm samples pooled from each plant. It should therefore be more practical and have wide applicability in mapping endosperm traits. The method was implemented by fitting the phenotypic means of endosperms into a Gaussian mixture model. Both the exact and approximate Expectation-Maximization algorithms were proposed to estimate the model parameters. The presence of the QTL was determined by likelihood ratio test statistics. Statistical power and other properties of the new method were investigated and compared to the current single-seed method under a variety of scenarios through simulation studies. The simulations suggest a reasonable sample size should be used to ensure reliable results. The proposed method was also applied to a simulated genome data for further evaluation. As an illustration, a real data of maize was analyzed to find the loci responsible for the popping expansion volume.

Endosperm genotyping as a strategy to differentiate the allele source in maize heterozygous progeny

“…This tissue originates from the fusion between a polar male gamete and two female gametophyte nuclei. This fusion generates the triploid endosperm nucleus, which develops by mitotic divisions, carrying two copies of the maternal alleles and one copy of the paternal allele in its genome (Wu et al, 2002;Wang et al, 2008). Therefore, for a locus with two alleles (M 1 and M 2 ), four genotypes (…”