Genome-Wide Association Mapping Using a Bayesian Mixture Model for Plant Height in Oryza sativa

Abstract

Genotypic and phenotypic data could be used to predict inheritance of complex traits for plant breeding in genome wide association mapping studies (GWAS). In GWAS using a single marker model may leads to suboptimal use of genotypic datasets. Alternatively, using whole genome, a Bayesian mixture model may cluster markers into predefined classes. We used 413 diverse accessions of Oryza sativa with 36900 Single Nucleotide Polymorphisms (SNPs) markers for plant height. We assumed different genetic architectures for the phenotype. We estimated genotypic heritability as 0.61. Bayesian mixture model detected 144, 446, 54 SNPs with explanatory levels of 0.0001, 0.001 and 0.01 respectively. Chromosome 1 (n=109), and 3 (n=85) had the highest explanatory genetic variances as 23% and 19%, respectively. Correlation between genomic predicted observations and actual observations was found to be 0.94. Since GWAS are mostly based on only one replication as was also the case in this study; results need to be confirmed by independent validation experiments.