QTL Mapping Based On Extra-High-Density Molecular Markers

Mapping of quantitative trait loci(QTLs)is a main approach in modern quantitative genetic studies.Traditional QTL mapping is mainly based on conventional molecular marker genetic maps.Conventional genetic maps usually have a low marker density,leaving a blank region between two adjacent markers.Therefore,testing of a QTL in the blank region can only be performed by linkage analysis according to the flanking markers.The classical interval mapping and composite interval mapping methods are all based on this principle.In recent ten or more years,because of the rapid development of highthroughput sequencing technology,high-density or extra-high-density genetic maps have been constructed in many species using high-throughput sequencing.There are basically no blank regions in high-density genetic maps.This means that it is unnecessary to use flanking markers but directly test individual markers to map QTLs.Hence,the QTL mapping methods for conventional genetic maps may not efficiently utilize the information of high-density maps.It is necessary to develop QTL mapping methods suitable for high-density maps.Another important development in QTL mapping study brought about by the high-throughput sequencing technology is genome-wide association study(GWAS)based on natural populations,which has become a common method for QTL mapping.However,the sensitivity and accuracy of GWAS depends on a large sample size.At present,although the cost of highthroughput sequencing has been dropped greatly,resequencing a large sample is still a huge cost for general researchers.Hence,how to increase the sensitivity and accuracy of GWAS without increasing the cost of resequencing is a problem demanding prompt solution.For the above problems,a statistical method suitable for QTL mapping based on high-density maps named composite marker test(CMT)was developed in this study.Meanwhile,an experimental design for GWAS was proposed,called population amplification by random cross(PARC).The main results are as follows:I.Composite marker test(CMT)1.The CMT method for DH and immortalized F2 populations was established.Its basic principle is: Every marker in the genome is tested through regression analysis by least squares and likelihood ratio test according to the additive-dominance genetic model.Multiple markers are included as cofactors in the model to control genetic background noise.After scanning the whole genome,QTLs are identified from LOD peaks.2.The method for screening CMT cofactors by stepwise regression according to the F statistic of the regression equation was established.How to reasonably select cofactors is always a problem not well solved in QTL mapping.In this study,it was found that the F statistic of the regression equation has maximum value,at which the regression model appears to be the optimal.Therefore,the markers included in the regression model with the maximum F value could be used as cofactors.Simulation study showed that the number of cofactors selected by this method is the best,which not only guarantees sufficiently large degrees of freedom for the regression equation,but is also able to well control background noise so as to the efficiency and accuracy of QTL mapping.3.A method to estimate LOD threshold for CMT by extrapolation from a small number of permutation tests was established.Permutation test is presently considered as the most reasonable method for estimating the significance threshold of QTLs.However,the method requires large amount of calculation,usually at least 1000 times.As a high-density map contains a great many markers,estimation of LOD threshold by permutation test will take a very long time,which is difficult to be conducted in practice.To handle this problem,a method was created in this study.In this method,only a small number(10-20 times)of permutation tests are needed.By dividing the genome and random assembling,the 5% significance thresholds of different genome lengths can be obtained.Therefore,the functional relationship between threshold and genome length can be found by curve regression.Then,according to the function,the LOD threshold for the genome being analyzed can be approximately estimated through extrapolation.4.Using simulated and real examples,the feasibility and efficiency of CMT was studied and compared with two traditional QTL mapping methods,composite interval mapping and inclusive composite interval mapping.The results indicated that CMT is an efficient QTL mapping method,which can increase the accuracy and power of mapping,better than the traditional methods for high-density maps.II.Population amplification by random cross(PARC)designThe PARC design was proposed for the GWAS based on a natural population of pure lines.Its basic idea is: from a basic population consisting of pure lines for GWAS,pure lines are allowed to be randomly crossed to produce a population containing a large number of hybrid lines.The genotype of each hybrid line can be deduced from its parental pure lines.Therefore,there is no need to resequence hybrid lines.It is only required to phenotype the hybrid line population for GWAS.For a basic population with n pure lines,the maximum number of hybrid lines can be up to n(n-1)/2,which is(n-1)/2 times as large as the basic population.Therefore,PARC design allows to get a very large population for GWAS with very small cost in resequencing.A simulation study was conducted using the SNP genotype data from rice.The results demonstrated the feasibility and efficiency of PRAC,which can dramatically increase the power and accuracy of QTL mapping by GWAS,and therefore is a very practical experimental design.