Discussion Of Several Issues Related To QTL Mapping Of Complex Traits In Crops

In recent years, with the development and promotion of molecular biology, quantitative genetics have made a raid progress.The research objective is the construction and application of gene/QTL mapping for quantitative traits by using DNA molecular markers and the final destination is to gradually establish molecular basis of quantitative genetics, so as to directly select quantitative genotype rather than to select quantitative phenotype in crop breeding. However, in aspects of crop gene/QTL mapping, there still exist a lot of related problems needed to further study, such as the results difference of different QTL mapping methods, the feasibility of detecting gene/QTL controlling general combining ability of backbone parants, the association between gene sequence variation by next generation sequencing technology and phenotypic traits, and the mechanism of how these gene/QTL impacted on one or more phenotypic traits collectively, these problems can be solved by using the methods of statistical genetics.In this paper, statistical analysis software R were used to study some main focus on quantitative genetics, such as similarities and differences between the results of different methods of QTL mapping, the application of Bayesian QTL mapping, the feasibility of QTL mapping for GCA-related, and the jointly genome selection procedure of multiple correlated traits under supersaturated situation, the main conclusions in this research are as follows:(1) R/qtl, which implemented by R language, is software especially for QTL analysis. In order to study the difference of analysis results among different mapping methods, plant height data of an F2:3maize populations were used to perform QTL mapping and compare the difference among interval mapping (IM), composite interval mapping (CIM), two-dimension scan and multiple-QTL fitness in mapping QTL for plant height. The results showed that a total of eight QTL were detected by all methods. In IM method, seven QTL were detected in sum, and the phenotype variation explained by them reached54.57%, three common QTL were detected by all four IM algorithms, and their locations, logarithm of the odds (LOD), confidence intervals and contribution of each QTL were almost the same. In CIM mapping, a total of five different QTL were detected, which accounted for about27.66%of phenotype variation, wherein two QTL located at158cM and222cM in chromosome3respectively could be detected by all four algorithms in CIM. The QTL located at36cM in chromosome7was found for the first time, and the parameters of other four QTL estimated by CIM were consistent with ones in IM methods. A total of seven QTL were detected using two-dimension scan, and they explained54.97%of phenotype variation. No interactive QTL were found. Except the QTL located at33.5cM in chromosome6, other QTL were consistent with those detected in IM method. Only three QTL were found by multiple-QTL fitness, and their total contribution to phenotype variation was up to21.52%. And still No interacting QTL were found by multiple-QTL fitness method. These results demonstrated that the QTL detected by above-mentioned methods were different in a certain extent not only at number but also at effects, whereas some main QTL could usually be detected by all methods. In addition, there were only QTL with main effect controlling plant height, no epistatic QTL were detected for this population.(2) Plant height is an important factor to corn yield. In order to explore the main effect, interactive effect and genetic architecture of plant height, an F2:3population was used to map QTL controlling plant height of maize by QTL mapping software package R/qtlbim which based on Bayesian statistics. With the one-dimension scan, a total of eight QTL were mapped on chromosome3to7, which the21ogBF of each QTL ranged from2.237to6.196. Two-dimension scan showed that a large number of interactive weeker signals existed, These interaction signals mainly concentrated on chromosome3to7. With the Bayes factor analysis, an optimal genetic architecture was obtained, and it was composed of six main effects QTL without interaction. The stepwise regression analysis showed that the genetic architecture was significant at0.001levels and could explain37.83%phenotype variation. Each of six QTL was very significant and could explain3.924%to10.776%of phenotype variation. This result indicated that the genetic architecture for plant height was relative simple. Interaction among QTL might be ignored because of its small effects.(3) General combining ability (GCA) is an important indicator for assessing parents. Studying the genetic basis of GCA and the feasibility of mapping QTL related with GCA (QTLgca) may provide a foundation in breeding. A NCII mating design, where a set of recombinant inbred lines (RIL) derived from two pure lines was used as the tested lines and several randomly selected pure lines were used as testers, and QTL mapping strategy was applied to investigate the genetic component of GCA, the impact factors for QTLgca mapping and the relationship between QTLqca and QTL controlling trait per se. Results indicated that when a trait was controlled by one pair of alleles, the GCA effect estimate of RIL and QTLgca mapping were all associated with additive effect, dominant effect of QTL controlling trait per se and the alleles frequency in testers. When a trait was controlled by two pairs of additive/dominance alleles, the GCA effect estimate of RIL and QTLgca mapping did not associate with whether the linkage between genes or not, and the impact factors of GCA estimate and QTLgca mapping were the same as that in one pair of alleles. When a trait was controlled by two pairs of interactive alleles, both GCA estimate and QTLgca mapping were all associated with the alleles frequency in testers, the major effect of QTL controlling trait and interactive effect between QTL. In addition, GCA estimate was also influenced by the recombination rate between loci. The allele’s frequency in testers and relative size of effects of QTL controlling trait per se were important factors not only to GCA estimation but also to QTLqca mapping, and the difference between QTLqca mapping and QTL mapping also depended on the genetic architecture of quantitative traits and analyzsis method of mapping QTL.(4) Measured traits are often related in traits-QTL association analysis; joint analysis of multiple related traits can improve the accuracy and precision of effect estimates, and can distinguish pleiotropy and close linkage. In this study, a supersaturated model with multiple related traits was simulated, partial least squares combined with two-stage variable selection approach wre adoped to reduce variable dimentions and the best model selection by adjusted AIC information criterion, and the results were compared with the single trait analysis. The results show that:the number of varieties, the total contribution rate, PIC and the size of candidate genes effects contributed significantly to the statistical power, the accuracy and precision of the estimate effect; compared with single trait analysis, multi-trait joint analysis can improve statistical power, increase the the accuracy and precision of estimated effect, significantly reduce the computation time and research costs; multi-trait joint analysis can also significantly improve the capability of detecting pleiotropic genes, even thought a gene had only impact on a single trait, joint analysis was also superior to the single analysis method; however for a candidate gene with minor PIC, the single trait analysis was better than joint nanlysis by partail least square, this maybe result from the loss of a variation of the variance in the principle component extraction in the process of partail least square regression. In conclusion, partial least squares analysis method has obvious advantages in joint variables selection and effects estimation for multiple correlated traits.