hide
Free keywords:
-
Abstract:
Objectives: Understanding the general genetic architecture (instead of identifying particular variants that are causal) of Arabidopsis thaliana traits based on full genomes. Methods: We sequenced ~240 Arabidopsis thaliana inbred lines with Illumina at deep coverage (ranging from 20 to 60) and measured multiple phenotypic traits like germination, genome size from flow cytometry, telomere length, as well as salt tolerance. Together with many published phenotypes including flowering time etc., we can compare the genetic architecture of different traits. To analyze the genetic architecture, we developed a linear mixed model in which there are two random terms that are assumed to be Gaussian with covariance matrixes as local and global kinship matrixes calculated by the variants detected from sequences. In contrast to the standard mixed model, we do not include fixed terms (except for the intercept). The rationale is that, hoping the global kinship term can guard against population structure, the local kinship term provides an estimate of phenotypic variance explained by the focal genomic region. This model is similar to local variance component analysis in Visscher’s lab (Yang et all Nat Genet 2010), however jointly accounting for the population structure. In our analysis, the relative contributions of local and global kinship terms to the heritability will be estimated by maximal likelihood ratio as well as a Bayesian approach. Results: Different traits perform quite differently in the genetic architecture analysis. Some are controlled by many small-effect regions and therefore can be assumed to be approximately infinitesimal; while some are contributed by a limited number of big-effect regions. Some traits do not show big difference when using SNPs only or indels included to calculate local kinship; while some do have more phenotypic variance explained when indels are considered. Some traits show genetic heterogeneity with multiple different variants contributing to the variance independently. Some traits are likely to be altered by gene-gene interactions. Conclusions: Sequence based whole genome analysis reveals important insights into genetic architecture of A thaliana.