Methodology For Linkage-map-based Quantitative Trait Loci Synthesis Analysis

Quantitative trait loci (QTL) mapping is to associate agronomic traits of interest with its QTL or gene, and marker assisted selection on the basis of the results from QTL mapping becomes an effective tool to improve the quality in crops and to accelerate crop breeding. On the other hand, comparative genomics, which make use of the information contained in the spatial organization of the genomes to infer the gene in the unknown genome, becomes a crucial tool to trace the evolution among species and to establish the functional regions of genomics.At present many QTL mapping approaches have been proposed to detect the effects of QTL, i.e., main, epistatic and QTL-by-environment interaction effects. In various genetic experiments, different QTL were detected for a same quantitative trait or some related traits by using various approaches. To utilize these identified QTL, it is necessary to carry out synthesis analysis. Its purpose is to get the refined QTL and to enhance the accuracy of the location. At present the most widely applied synthesis analysis is meta-analysis. However, the key in the meta-analysis of QTL is to have a core marker and the simplify adjunction of QTL close to a core marker will bias the final results in the meta analysis of QTL by estimates of the position and the confidence intervals of the original QTL. And the meta-analysis is just a computational method that is not supported by biological mechanism. In addition, the Akaike's information criteria (AIC) used in meta-analysis has been confirmed to have a low power in the selection of model. Therefore, the shortcomings are strong enough to justify seeking improvements to synthesis analysis.The purpose of synthesis analysis of QTL is to obtain some common results in various genetic experiments with various mapping approaches. Obviously, the marker are the key to link the phenotypic values of quantitative trait to gene in the QTL mapping. This indicates that the homology relationship among markers could be employed for the synthesis analysis. The proposed methods for homology segment identification in this paper are based on physical map and final results are inferred on statistics. This limits the comparison among linkage groups. In addition, the rule used in the conventional methods are empirical. Therefore, a new method of homology segment identification based on recombination frequency was proposed in this paper. The homology segments, deduced from the molecular markers, could be utilized to aggregate QTL and to exclude QTL even if they share same markers. However, the obtained information is complicated, the imbalanced and heterogeneity data may bias the results in the conventional classification method. Therefore, a new one-class classification approach is needed. At last, a real data analysis in cotton is carried out by incorporating the new one-class classification approach and homology segment identification method.The results are as follows.1) Based on Bayesian rough set model, a new one-class classification approach, with the characteristics of both rough set and one-class classification, is proposed in this article. The new method is to use minimum model assumptions and to admit the ignorance when the data quality is low. And the new method here is available in the analysis of the imbalanced and/or heterogeneity data. Four simulated datasets and two real datasets were used to validate this new method. Compared to other decision tree methods and dominance-based rough set approach, the attributes identified by the new method shows higher reliability. In additions, the indicators for measuring the data quality, including the optimal sample fraction, the discernability and the accuracy, were proposed in this paper.2) The biology foundation for the homology segment identification method, which is based on both graph theory and pattern identification, is the recombination fraction during meiosis. Compared to other identification methods, the new method has its particular characteristics, i.e., the spatial conservation and sensitive to the change of chromosome. In addition, the obtained pattern linkage graph, layer matrix and synteny block are used to study the exchanged behavior of the meiosis and the evolution relationship among chromosomes.3) Using the data of six linkage groups from three independent experiments in cotton, the meta-analysis and the evaluation of the relationship between the A and D sub-genomes were carried out. The results showed that the layer matrix in this essay could recognize the relationship between the A and D sub-genomes, and the synteny segments not only can establish QTL cluster but also can be used to find out the homology relationship of QTL. In addition, a new one-class classification approach implemented in the real data analysis above can not only recover the innate complex of biology mechanism but also identify the hidden rules, which could be used to support marker assisted selection (MAS) and map-based cloning.