Fine-mapping quantitative trait loci with a medium density marker panel: efficiency of population structures and comparison of linkage disequilibrium linkage analysis models

Abstract

SummaryRecently, a Haley–Knott-type regression method using combined linkage disequilibrium and linkage analyses (LDLA) was proposed to map quantitative trait loci (QTLs). Chromosome of 5 and 25 cM with 0·25 and 0·05 cM, respectively, between markers were simulated. The differences between the LDLA approaches with regard to QTL position accuracy were very limited, with a significantly better mean square error (MSE) with the LDLA regression (LDLA_reg) in sparse map cases; the contrary was observed, but not significantly, in dense map situations. The computing time required for the LDLA variance components (LDLA_vc) model was much higher than the LDLA_reg model. The precision of QTL position estimation was compared for four numbers of half-sib families, four different family sizes and two experimental designs (half-sibs, and full- and half-sibs). Regarding the number of families, MSE values were lowest for 15 or 50 half-sib families, differences not being significant. We observed that the greater the number of progenies per sire, the more accurate the QTL position. However, for a fixed population size, reducing the number of families (e.g. using a small number of large full-sib families) could lead to less accuracy of estimated QTL position.

Mapping genomic regions affecting milk traits in Sarda sheep by using the OvineSNP50 Beadchip and principal components to perform combined linkage and linkage disequilibrium analysis

“…The proposed approach allowed us to solve the model by the least square method, which avoid a computational expensive variance component analysis. The advantages of approaches based on LDLA regression versus those on variance components, in term of ease of use and computing time, are well known [29] and have been clearly demonstrated by Roldand et al [52].…”

Identification of quantitative trait loci underlying milk traits in Spanish dairy sheep using linkage plus combined linkage disequilibrium and linkage analysis approaches

“…Additionally, an optimal number of half-sib families should be included in the experiment. In a simulation study, Roldan et al (2012) demonstrated that the analysis of 15 to 50 mediumsized half-sib families (between 20 and 65 daughters per sire) produced more accurate QTL mapping results than a small number of large families (5 sires with 200 progeny each) or a large number of small half-sib families (100 sires with 10 progeny each).…”

Using haplotypes for the prediction of allelic identity to fine-map QTL: characterization and properties

“…The statistical test used may also favor some AIP, such as IBS hap , IBS m and Beagle, over others due to numerical stability when estimating variance components. As such, solving mixed model equations when covariance matrices are close to singularity due to AIP computation has been reported as an issue, and clustering strategies for haplotypes, which actually modify the properties of the AIP matrices, have been proposed to facilitate computation [ 42 , 43 ]. The major drawback of the matrix distance approach is related to this advantage: a particularly efficient AIP or a particularly efficient haplotype size, identified from the matrix distance, that can not be used in association studies would be of no value.…”

“…Yet advantages have been reported for P(IBD) compared to IBS hap in some situations. For example, Roldan et al [ 43 ] showed better accuracy for P(IBD) compared to IBS hap , after a clustering step for haplotypes when marker intervals were equal to 0.05 cM between SNPs, but not when they reached 0.25 cM. However in Roldan et al [ 43 ], different statistical models were applied to P(IBD) versus IBS hap (mixed model versus fixed effects model respectively).…”