Crape myrtle(Lagerstroemia indica)is a globally used ornamental woody plant and is the representative species of Lagerstroemia.However,studies on the evolution and genomic breeding of L.indica have been hindered by th...Crape myrtle(Lagerstroemia indica)is a globally used ornamental woody plant and is the representative species of Lagerstroemia.However,studies on the evolution and genomic breeding of L.indica have been hindered by the lack of a reference genome.Here we assembled the first high-quality genome of L.indica using PacBio combined with Hi-C scaffolding to anchor the 329.14-Mb genome assembly into 24 pseudochromosomes.We detected a previously undescribed independent whole-genome triplication event occurring 35.5 million years ago in L.indica following its divergence from Punica granatum.After resequencing 73 accessions of Lagerstroemia,the main parents of modern crape myrtle cultivars were found to be L.indica and L.fauriei.During the process of domestication,genetic diversity tended to decrease in many plants,but this was not observed in L.indica.We constructed a high-density genetic linkage map with an average map distance of 0.33 cM.Furthermore,we integrated the results of quantitative trait locus(QTL)using genetic mapping and bulk segregant analysis(BSA),revealing that the major-effect interval controlling internode length(IL)is located on chr1,which contains CDL15,CRG98,and GID1b1 associated with the phytohormone pathways.Analysis of gene expression of the red,purple,and white flower-colour flavonoid pathways revealed that differential expression of multiple genes determined the flower colour of L.indica,with white flowers having the lowest gene expression.In addition,BSA of purple-and green-leaved individuals of populations of L.indica was performed,and the leaf colour loci were mapped to chr12 and chr17.Within these intervals,we identified MYB35,NCED,and KAS1.Our genome assembly provided a foundation for investigating the evolution,population structure,and differentiation of Myrtaceae species and accelerating the molecular breeding of L.indica.展开更多
联邦学习和群智学习等协作学习技术,能够在保护数据隐私的条件下充分利用分布在各地的分布式数据深度挖掘数据中所蕴含的知识,拥有非常广阔的应用前景,尤其是在强调隐私惯例和道德约束的医疗健康领域.任何协作工作都需要选择可靠的参与...联邦学习和群智学习等协作学习技术,能够在保护数据隐私的条件下充分利用分布在各地的分布式数据深度挖掘数据中所蕴含的知识,拥有非常广阔的应用前景,尤其是在强调隐私惯例和道德约束的医疗健康领域.任何协作工作都需要选择可靠的参与方,协作学习中全局模型的性能在很大程度上取决于参与方的选择.然而,现有研究在选择参与方时都没有对不同机构医疗数据中存在的异质性加以直接关注.导致包含稳定性在内的全局模型的性能难以得到保障.提出了从信誉的角度尝试探索求解该问题.以迭代协作学习的方式尽可能挑选出具有良好信誉的参与方进行协作学习,以获得稳定可靠的高性能全局模型.首先,提出了一个描述医疗机构数据质量的AI信誉值指标AMP(AI medical promise),以帮助其在医疗领域中形成良好的AI生态.其次,建立了一个基于后向选择的迭代协作学习(colback-learning)框架.在单次协作学习任务中,利用后向选择方法以多项式时间复杂度迭代计算出性能良好且稳定的全局模型,完成AMP计算和积累.在AMP信誉值计算中,制定了一个综合考虑全局性能指标的评分函数,以针对医疗领域更有效地指导全局模型的训练.最后,使用真实医疗数据模拟多样化的协作学习场景.实验表明,colback-learning能够选择可靠参与方训练得到性能良好的全局模型,模型的性能稳定性比现有最好的参与方选择方法提高了1.3~6倍.全局模型的可解释性与集中式学习保持了较高的一致性.展开更多
基金supported by National Key R&D Program of China(2019YFD1001004,2019YFD1000402)the program for Science and Technology of Beijing(Z181100002418006)the Special Fund for Beijing Common Construction Project.
文摘Crape myrtle(Lagerstroemia indica)is a globally used ornamental woody plant and is the representative species of Lagerstroemia.However,studies on the evolution and genomic breeding of L.indica have been hindered by the lack of a reference genome.Here we assembled the first high-quality genome of L.indica using PacBio combined with Hi-C scaffolding to anchor the 329.14-Mb genome assembly into 24 pseudochromosomes.We detected a previously undescribed independent whole-genome triplication event occurring 35.5 million years ago in L.indica following its divergence from Punica granatum.After resequencing 73 accessions of Lagerstroemia,the main parents of modern crape myrtle cultivars were found to be L.indica and L.fauriei.During the process of domestication,genetic diversity tended to decrease in many plants,but this was not observed in L.indica.We constructed a high-density genetic linkage map with an average map distance of 0.33 cM.Furthermore,we integrated the results of quantitative trait locus(QTL)using genetic mapping and bulk segregant analysis(BSA),revealing that the major-effect interval controlling internode length(IL)is located on chr1,which contains CDL15,CRG98,and GID1b1 associated with the phytohormone pathways.Analysis of gene expression of the red,purple,and white flower-colour flavonoid pathways revealed that differential expression of multiple genes determined the flower colour of L.indica,with white flowers having the lowest gene expression.In addition,BSA of purple-and green-leaved individuals of populations of L.indica was performed,and the leaf colour loci were mapped to chr12 and chr17.Within these intervals,we identified MYB35,NCED,and KAS1.Our genome assembly provided a foundation for investigating the evolution,population structure,and differentiation of Myrtaceae species and accelerating the molecular breeding of L.indica.
文摘联邦学习和群智学习等协作学习技术,能够在保护数据隐私的条件下充分利用分布在各地的分布式数据深度挖掘数据中所蕴含的知识,拥有非常广阔的应用前景,尤其是在强调隐私惯例和道德约束的医疗健康领域.任何协作工作都需要选择可靠的参与方,协作学习中全局模型的性能在很大程度上取决于参与方的选择.然而,现有研究在选择参与方时都没有对不同机构医疗数据中存在的异质性加以直接关注.导致包含稳定性在内的全局模型的性能难以得到保障.提出了从信誉的角度尝试探索求解该问题.以迭代协作学习的方式尽可能挑选出具有良好信誉的参与方进行协作学习,以获得稳定可靠的高性能全局模型.首先,提出了一个描述医疗机构数据质量的AI信誉值指标AMP(AI medical promise),以帮助其在医疗领域中形成良好的AI生态.其次,建立了一个基于后向选择的迭代协作学习(colback-learning)框架.在单次协作学习任务中,利用后向选择方法以多项式时间复杂度迭代计算出性能良好且稳定的全局模型,完成AMP计算和积累.在AMP信誉值计算中,制定了一个综合考虑全局性能指标的评分函数,以针对医疗领域更有效地指导全局模型的训练.最后,使用真实医疗数据模拟多样化的协作学习场景.实验表明,colback-learning能够选择可靠参与方训练得到性能良好的全局模型,模型的性能稳定性比现有最好的参与方选择方法提高了1.3~6倍.全局模型的可解释性与集中式学习保持了较高的一致性.
