Models for marker-assisted genetic evaluation with multiple QTL in a crossbred population

The efficiency of alternative models for marker-assisted genetic evaluation with multiple previously identified QTL for a trait with heritability 0.1 was evaluated by stochastic simulation. Three biallelic unlinked additive QTL were simulated in the middle of marker intervals of 0, 10, and 20 cM, with each QTL explaining 12, 6, or 3% of genetic variance in the F2 of a cross between inbred lines. Three models for marker-assisted genetic evaluation were compared to standard BLUP (B): BM = B with fixed marker effects; BMR = BM plus inclusion of random QTL effects; M = selection on the number of favorable marker alleles. All MAS models resulted in greater responses than B in initial generations, but extra gains declined over generations. The impact of the magnitude of QTL variance used for genetic evaluation for BMR on average QTL frequencies and response was limited. Selection with M gave greater response than B only up to the F5. For BM and BMR, extra response over B and QTL frequencies increased when QTL effects increased and size of marker intervals decreased. The number of QTL that explained a given total amount of variance had no effect on the ranking of models in terms of QTL frequencies although a larger number of QTL resulted in higher genetic gains in later generations. Heritability had no effect on the ranking of the models. Based on genetic gains and ease of implementation, model BM is recommended as the most suitable model for marker-assisted selection in crosses of inbred lines.