Bayesian estimation of multi-allele QTLs for agricultural traits in tomato using recombinant inbred lines derived from two F 1 hybrid cultivars

In comparison with a population derived from bi-parental crosses of pure lines, a customized experimental population derived from multi-parental crosses offers a significant advantage for detecting multiple QTL alleles. We established a population of recombinant inbred lines (RILs) derived from crossing two commercial elite F 1 hybrid cultivars of tomato (G 1 generation). This population can be considered four-way RILs because it originates from four parental inbred lines (G 0 ), i.e., four parents of the F 1 cultivars. We applied two different Bayesian mapping methods. One was based on the Markov chain Monte Carlo (MCMC) algorithm, and the other was a newly developed rapid method using variational approximation to detect QTLs for agricultural traits, including yield and soluble solid content. Each plant from the resulting population (G 1 F 1 ; 240 plants) was used as the basis for repeated cycles of selfing through single-seed descent to obtain a RIL population (G 1 F t ( t ≥ 6), n = 206). Assuming that there were four possible alleles at each QTL derived from the founders, we inferred marker haplotypes of the G 1 cultivars and used the inferred haplotypes to obtain the QTL genotypes of the RILs. We investigated the segregation of markers transmitted from the hybrid cultivars to the RILs and constructed a linkage map, 1221.8 cM in length. Based on the linkage map, 55 significant QTLs were detected for the analyzed traits. Considering the effects of the four types of QTL alleles estimated in this study, we present a practical approach to genomic selection to improve trade-off traits.