The Genetic Diversity Of White Maize Landraces And Promary Research Of Association Analysis In F₁ Families

Maize (Zea mays.L) is the most economically important crops in the world. Along with the improvement of the economic, it is necessary to develop genetic breeding research of maize to meet the requirement of market demand. Genetic diversity is the basic portion of biological diversity and is the base of species diversity. Genetic diversity of maize (Zea mays.L) plays a key role for breeding progress. Association analysis is a powerful method to detect and locate quantitative trait loci (QTL) based on the strength of the correlation between genetic markers and a trait. Recent advances in genomic technology and development of robust statistical analysis methods make association mapping appealing and affordable to plant research programs.Analysis of germplasm resource distribution and genetic diversity were conducted by using 50 white maize landraces in Southwest China, and two maize landraces with one comes from Yunnan province and the other comes from Guangxi Zhuang Autonomous Region were used in this study. A finite number of individuals were randomly selected from the two maize landraces as female and male parents to make pairwise crosses and generate 135 F1 families, including 70 positive cross and 65 negative cross genealogies. Association analysis was conducted in 135 F1 families using 51 SSR markers, and the efficiency of this method was evaluated through simulation. The major results were as follows:1. White maize landraces included in the study were evaluated for genetic diversity and germplasm resource distribution based on 51 SSR with Phenotypc and SSR data. These maize germplasms varied significantly in all of the measured morphological traits. Principal components cluster analysis showed that these accessions originated from 7 populations, and most of these landraces were clustered into one group. However, a few varieties could be clustered into another single group. Bulk DNA samples were tested to make genetic diversity analysis. A Total of 515 alleles were scored, averaging 12 and ranging from 5 to 19 alleles per locus. Genetic similarity coefficients ranged from 0.574 to 0.84, with an average of 0.885. On the basis of the genetic similarity coefficients, clustering analysis separated these landraces into 8 groups.2. Association analysis between agronomic traits and SSR marker was made between 70 positive cross genealogies,65 negative cross genealogies,135 Fi families, and F1 families with the effects of reciprocal cross excluded respectively. Association mapping was conducted by the GLM model and K model of TASSEL software. The results indicated that a total of 14 common significant loci were identified by two different models of association analysis in the F1 families on the level of 0.001, including 2,6,2,4 significant loci respectively. Obviously, more stable and significant markers were found in these F1 families with the effects of reciprocal cross excluded.3. Four different models including GLM, Q model, K model, and Q+K model were used to detect the significant markers associated with seven morphological traits in TASSEL, with minor allele frequencies were exclude. The results indicated that a total of six common significant loci were identified by these different models of association analysis in the F1 families at p<0.01, and only one significant marker association with SL was identified at p＜O.OOO1. Quantile-quantile plots were made to evaluate the model fit for different traits, as a result, the model combined population structure and relative kinship (Q+K) performed better than the model only controlling relative kinship (K) and also only controlling population structure (Q).4. The results of association analysis using F1 families with phenotypic values were corrected by reciprocal cross effects were used to set simulated parameter. A total of 100 F1 families were obtained after running the program QU-GENE and QuLine, and two models (GLM and K) were considered. Across 100 simulation, the average power of all QTL detection for the seven traits using the two models were 88.64% and 83.64% respectively, and the number of false QTL were decreased from 399 in GLM models to 199 in K models.