图1 几种植物基因组Illumina测序数据K-mer分布曲线 a:测序错误导致的峰,深度只有1-2;b:单倍体或者纯合二倍体基因组的主峰;c:低拷贝数重复序列组成的峰,深度常为主峰的2倍;d:高频重复序列组成的峰。在杂合二倍体基因组中,b1峰包含杂合区域的k-mer,b2峰包含纯合区域的k-mer。b1深度只有平均深度的一半。在杂合同源四倍体植物中,b1和b2峰都表示杂合区域的k-mer,b3峰表示纯合区域的k-mer。该图根据文献[4,5,6,7,8,9]绘制

Fig. 1 K-mer volume histograms of illumina sequencing data from several plant genomes a:The first peak,at the depth at 1-2,is derived from sequencing error. b:The second peak,is the main peak in haploid genome or homozygous diploid genome. c-d:The peak c,is composed of the repetitive k-mers with relative low copy number while the right-most peak d,is composed of the highly repetitive k-mers. In heterozygous diploid genome,the peak b1 contains k-mers derived from heterozygous regions and the peak b2 contains k-mers derived from homozygous regions. The depth of peak b1 is only half of the average sequencing depth. In heterozygous autotetraploid genome,both of the peak b1 and b2 present heterozygous k-mers and the peak b3 presents homozygous k-mers. This figure is modified based on the references[4,5,6,7,8,9]

图2 三种植物基因组组装和分型方案 a:基于“亲本-子代”测序的分型方案Triobin[41]。利用亲本测序数据的特异性K-mer将子代的测序数据分成两份,分别组装出两个亲本的单体型。b:基于单倍体群体测序的分型方案[44]。预先组装的BAC片段作为分型的输入序列。研究人员测序了12个花粉细胞,并开发barcoding的方法将BAC片段的基因型转换成12位的二进制条码。该方法中的BAC序列可以替换成HiFi read或组装的contig等高准确率长片段。c:基于自交分离群体的分型方案[7]。该方案从头组装出二倍体contig,并测序分离群体对contig进行基因型鉴定。构建遗传图谱区分出不同的染色体,再利用基因型的相似性区分同一染色体、不同单体型的contig

Fig. 2 Three assembly and genotyping strategies of genome in plants a:The trio-sequencing based genotyping strategy Triobin[41]. The sequencing reads of F1 hybrid was firstly partitioned into paternal and maternal sets using the parental unique K-mers,then assembled separately into two haplotypes. b:Genotyping based on haploid population[44]. To phase the pre-assembled BAC clones,12 pollen cells were sequenced individually and the alignment results between each BAC and each pollen cell were encoded into a 12-bit binary barcode. In this approach,the BAC clones could be replaced with the high-accuracy and long segment such as HiFi reads or assembled contigs,etc. c:Genotyping based on selfing segregation population[7]. Firstly,the diploid contigs were de novo assembled,and a segregation population was sequenced to identify the genotypes of contigs. Then,the genetic map was constructed to identify different chromosomes. Lastly,the contigs belonging to same chromosome were partitioned into different haplotypes based on their genotypes’ similarity

图3 泛基因组构建的三种方式 a:迭代组装泛基因组。通过将序列比对回参考基因组,提取未比对序列进行组装,迭代延长参考基因组构建泛基因组;b:从头组装构建泛基因组。对所有个体进行从头组装和注释,通过基因聚类算法构建泛基因集合,根据基因在各品系中出现的频率进行分类,得到核心基因集和可变基因集,根据线性模型绘制泛基因组累积曲线图;c:图基因组。基于参考基因组进行变异提取,整合变异数据集进行图基因组构建,灰框展示不同于参考基因组的路径,右图展示图基因组两个区域的真实图形结构

Fig. 3 Three approaches of assembling pan-genome a: Iterative assembly pan-genome. Construction of pan-genome by mapping reads to a reference genome,the unaligned reads were then assembled into novel contigs and they were iteratively added to the reference genome. b: De novo assembly pan-genome. Multiple genomes were assembled and annotated,and pan-gene clusters were predicted using clustering algorithm. Gene-clusters were further cataloged to core- and dispensable-sets according to the cluster frequency among all samples. The pan- and core-genome curves were fitted using nonlinear models. c: Graph-based genome. A pan-genome graph can be constructed by integrating variations to a reference genome. Grey box indicated the alternative path differing from reference genome. Right side showed the actual structure of 2 regions in graph-based genome