Additional file 2 of Improving genomic prediction of vitamin C content in spinach using GWAS-derived markers

Supplementary Material 2: Figure S1. Distribution of the 147,977 single nucleotide polymorphism (SNP) markers on the 6 chromosomes of spinach. Spinach chromosomes are on the vertical axis. Chromosome length in Mb is on the horizontal axis, and the color represents the number of SNPs per 1 Mb window size, SNP density. Figure S2. |Population structure and phylogeny analysis of 347 spinach accessions. A. Population stratification into three population groups based on ΔK analysis. Two (SQ1 and SQ2) and three (SQ1, SQ2 and SQ3) populations estimated from the admixture model (horizontal axis denotes accessions and vertical axis denotes probability values). B. Phylogenetic tree of two and three sub-populations by neighbor-joining (NJ) method drawn by GAPIT 3 in 347 spinach lines, where PCA = 2 (left) and PCA = 3 (right). Figure S3. The Manhattan and quantile–quantile (QQ) plots of BLINK, FarmCPU, and MLM models for ascorbic acid (AsA) content by GAPIT 3. The x-axis and y-axis in the Manhattan plots denote log10 ( P ) values and identified SNPs on spinach chromosomes. And x-axis and y-axis in the QQ plots denotes observed and expected log10 ( P ) values. Figure S4. The Manhattan and QQ-plot of SMR, GLM, and MLM models for ascorbic acid (AsA) content by TASSEL 5. The x-axis and y-axis in the Manhattan plots denote log10 ( P ) values and identified SNPs on spinach chromosomes. And x-axis and y-axis in the QQ plots denotes observed and expected log10 ( P ) values. Figure S5. Display of LD plots with SNPs on haplotype blocks in different chromosomes of spinach genome. A. Chr. 1, B. Chr. 2, C. Chr. 3, D. Chr. 4, E. Chr. 5, and F. Chr. 6. The triangle (black color) on respective chromosome represents different blocks with VC related SNP markers. Figure S6. Genomic prediction for ascorbic acid (AsA) content in 347 spinach lines, where (top) fourteen different SNP number sets from 12 randomly selected SNPs to 60,000 SNPs and two GWAS SNP marker sets estimated by seven GP models. A. Genomic prediction of Bayes A (BA), Bayes B (BB), Bayes LASSO (BL), and Bayes ridge regression (BRR); B. Genomic prediction of Random forests (RF), ridge-regression best linear unbiased prediction (rrBLUP) and Support vector machines (SVM). C. Genomic prediction of BRR model. Y-axis shows PA (r-value) and X-axis shows GP models in all the plots, and D. Genomic prediction using seven GP models using r1000 and m12 and m62 SNP sets

补充材料2：图S1. 147,977个单核苷酸多态性（single nucleotide polymorphism, SNP）标记在菠菜6条染色体上的分布。纵轴为菠菜染色体，横轴为染色体长度（单位：Mb），颜色代表每1 Mb窗口内的SNP数量（即SNP密度）。图S2. 347份菠菜种质的群体结构与系统发育分析。A. 基于ΔK分析的群体分层结果，将群体划分为3个组。混合模型估计得到2个群体（SQ1和SQ2）及3个群体（SQ1、SQ2和SQ3）（横轴为种质，纵轴为概率值）。B. 利用GAPIT 3软件通过邻接（neighbor-joining, NJ）法构建的347份菠菜材料的2个和3个亚群系统发育树（左为PCA=2，右为PCA=3）。图S3. 利用GAPIT 3软件得到的抗坏血酸（ascorbic acid, AsA）含量的BLINK、FarmCPU和MLM模型的曼哈顿图与分位数-分位数（quantile-quantile, QQ）图。曼哈顿图中横轴为log10(P)值，纵轴为菠菜染色体上的已鉴定SNP；QQ图中横轴为观测log10(P)值，纵轴为期望log10(P)值。图S4. 利用TASSEL 5软件得到的抗坏血酸（AsA）含量的SMR、GLM和MLM模型的曼哈顿图与QQ图。曼哈顿图中横轴为log10(P)值，纵轴为菠菜染色体上的已鉴定SNP；QQ图中横轴为观测log10(P)值，纵轴为期望log10(P)值。图S5. 菠菜基因组不同染色体上SNP的单倍型块连锁不平衡（linkage disequilibrium, LD）图展示。A. 1号染色体，B.2号染色体，C.3号染色体，D.4号染色体，E.5号染色体，F.6号染色体。各染色体上的黑色三角形代表含有VC相关SNP标记的不同单倍型块。图S6. 347份菠菜材料抗坏血酸（AsA）含量的基因组预测，其中（顶部）14个不同的SNP数量集（从12个随机选择的SNP到60000个SNP）及2个GWAS SNP标记集通过7个基因组预测（genomic prediction, GP）模型进行估计。A. Bayes A（BA）、Bayes B（BB）、Bayes LASSO（BL）和贝叶斯岭回归（Bayes ridge regression, BRR）的基因组预测结果；B. 随机森林（Random forests, RF）、ridge回归最佳线性无偏预测（ridge-regression best linear unbiased prediction, rrBLUP）和支持向量机（Support vector machines, SVM）的基因组预测结果；C. BRR模型的基因组预测结果。所有图中纵轴为预测精度（prediction accuracy, PA）（r值），横轴为GP模型；D. 利用r1000、m12和m62 SNP集通过7个GP模型进行的基因组预测结果。