UMR AGAP, CIRAD, Montpellier, France.
PLoS One. 2012;7(3):e33470. doi: 10.1371/journal.pone.0033470. Epub 2012 Mar 13.
Population structure, extent of linkage disequilibrium (LD) as well as signatures of selection were investigated in sorghum using a core sample representative of worldwide diversity. A total of 177 accessions were genotyped with 1122 informative physically anchored DArT markers. The properties of DArTs to describe sorghum genetic structure were compared to those of SSRs and of previously published RFLP markers. Model-based (STRUCTURE software) and Neighbor-Joining diversity analyses led to the identification of 6 groups and confirmed previous evolutionary hypotheses. Results were globally consistent between the different marker systems. However, DArTs appeared more robust in terms of data resolution and bayesian group assignment. Whole genome linkage disequilibrium as measured by mean r(2) decreased from 0.18 (between 0 to 10 kb) to 0.03 (between 100 kb to 1 Mb), stabilizing at 0.03 after 1 Mb. Effects on LD estimations of sample size and genetic structure were tested using i. random sampling, ii. the Maximum Length SubTree algorithm (MLST), and iii. structure groups. Optimizing population composition by the MLST reduced the biases in small samples and seemed to be an efficient way of selecting samples to make the best use of LD as a genome mapping approach in structured populations. These results also suggested that more than 100,000 markers may be required to perform genome-wide association studies in collections covering worldwide sorghum diversity. Analysis of DArT markers differentiation between the identified genetic groups pointed out outlier loci potentially linked to genes controlling traits of interest, including disease resistance genes for which evidence of selection had already been reported. In addition, evidence of selection near a homologous locus of FAR1 concurred with sorghum phenotypic diversity for sensitivity to photoperiod.
利用代表世界范围内多样性的核心样本，对高粱进行了群体结构、连锁不平衡（LD）程度以及选择特征的研究。总共对 177 个品种进行了基因型分析，使用了 1122 个具有物理锚定的 DArT 标记。比较了 DArT 对高粱遗传结构的描述特性与 SSR 和先前发表的 RFLP 标记的描述特性。基于模型（STRUCTURE 软件）和邻接法多样性分析，确定了 6 个组，证实了先前的进化假设。不同标记系统之间的结果基本一致。然而，DArT 在数据分辨率和贝叶斯分组分配方面更为稳健。用平均 r(2)衡量的全基因组连锁不平衡从 0.18（0 到 10kb 之间）下降到 0.03（100kb 到 1Mb 之间），在 1Mb 之后稳定在 0.03。使用 i.随机抽样、ii.最大长度子树算法（MLST）和 iii.结构组来测试样本量和遗传结构对 LD 估计的影响。通过 MLST 优化群体组成可以减少小样本中的偏差，并且似乎是选择样本以充分利用 LD 作为结构群体基因组图谱方法的有效方法。这些结果还表明，在涵盖世界范围内高粱多样性的收藏中，可能需要超过 10 万个标记才能进行全基因组关联研究。对鉴定出的遗传群体之间的 DArT 标记分化进行分析，指出了潜在与控制感兴趣性状的基因连锁的异常基因座，包括已报道存在选择证据的抗病基因。此外，在 FAR1 同源基因座附近的选择证据与高粱对光周期的敏感性表型多样性一致。
