Extended Data Fig. 3: Population structure analysis and phenotypic change within four subgroups during rice hybrid improvement breeding.

a, Admixture analysis with K = 4. Samples were clustered according to their cytoplasm types. The bar on the bottom represents cytoplasm types. The half-height bar represents cytoplasm types from two-line breeding systems, comprising Twoline-Jap and Twoline-Ind organelle types. The full-height bar represents cytoplasm types from three-line breeding system, comprising Wild abortive and Boro II organelle types. b, Kinship relationships for hybrids within three breeding periods. A circle indicats a hybrid accession, which is colored according to cytoplasmic type. Pairs of accessions sharing close relationship (kinship coefficients≥0.9) were connected by beelines. c, Phenotypic change of hybrids within four subgroups during improvement breeding. Sample size is shown in parenthesis. For WA subgroups, significant test was conducted by one-way ANOVA for data with homoscedasticity distribution or Kruskal-Wallis test for data with heteroscedasticity distribution. And multiple comparisons were further conducted using the least significant difference (LSD) method with ‘Bonferroni’ correction for homoscedasticity distribution or the Nemenyi test for heteroscedasticity distribution. Different lowercase letters above the boxplots indicate significant phenotype difference (p ≤ 0.05). For TJ, TI and BT subgroups, significant test was conducted by two-sided t-test for data with homoscedasticity distribution or two-sided wilcoxon test for data with heteroscedasticity distribution.