Estimation Of Recombination Frequency In Bi-Parental Genetic Populations And Properties Of The Test Statistic Under Null Hypothesis In Quantatitive Trait Loci (QTL) Mapping

Many traits in crop breeding are controlled by quantitiative trait loci (QTL). With thedevelopment of molecular markers, linkage analysis has become more widely in genetic studies ofthese traits. Precisely estimating recombination frequency and selecting appropriate LOD threshold inQTL mapping studies are two basic problems in linkage analysis. They are fundamental for finemapping, marker-assisted backcrossing, marker-assisted selection and map-based cloning.In this study we compared the linkage LOD score, the deviation between estimated and realrecombination frequency and their respective standard error in12bi-parental genetic populations byextensive simulations in the case that two marker loci are both co-dominant. We also compared thetheoretical standard error, theoretical population size required to make at least one recombinanthappen and to declare the significance linkage between two markers (LOD≥3) by theoreticalderivation. Results indicated that (1) tighter linkage (smaller value of recombination frequency) andlarger population size resulted in higher detect power and more accurate recombination frequencyestimation;(2) repeated backcross populations were not ideal for recombination frequency estimationbecause backcrossing leads to the allele frequency apart from0.5;(3) with co-dominant markers, F2and F3populations showed advantages on recombination frequency estimation, as they contain all thepossible nine genotypes, so more recombinant information is contained;(4) larger population size isneeded to make at least one recombinant happen and to declare the linkage between two loci whenusing dominant and recessive markers, so dominant and recessive marker were not favorable inrecombination frequency estimation.We investigated the properties of LRT statistic of one scanning point under null hypothesis in QTLmapping, the factors affecting the cumulative distribution of maximum LOD score, and the relationshipbetween the effective number of independent tests and the length of chromosome by simulation method.Results indicated that (1) the LRT statistic of one-dimensional scanning of additive-dominant QTL andtwo-dimensional scanning of epistatic QTL followed chi-square distribution, and the degree of freedom(df) was equal to the number of genetic parameters to be estimated;(2) the number of chromosome,population size and error variance did not have any effects on the distribution of test statistic under nullhypothesis, that meant having no impacts on the selection of LOD threshold, while population type,genome size and marker density had significant impacts, and higher marker density and longerchromosome resulted in higher LOD threshold;(3) the effective number of independent tests (Meff) wasproportional to the length of chromosome in one-dimensional scanning of additive-dominant QTL, andMeffwas the squared relationship to the length of chromosome in two-dimensional scanning of epistaticQTL. With the help of Bonferroni correction, we could acquire the relationship between point-wise andgenome-wide significance levels. Therefore, it’s easy to calculate the threshold at given genome-widesignificance level if we know the population type, marker density, genome size.