Phylogenomic analysis
To construct a phylogenetic tree, we used the predicted protein files from the S. rosmarinus genome and 23 other species (Antirrhinum majus , Arabidopsis thaliana , Daucus carota , Boea hygrometrica , Coffea canephora ,Amborella trichopoda , Erythranthe guttata , Sesamum indicum , Scutellaria baicalensis , Glycine max ,Handroanthus impetiginosus , Olea europaea , Oryza sativa , Populus trichocarpa , Beta vulgaris , Striga asiatica , Solanum lycopersicum , Vitis vinifera , Zea mays , Ocimum tenuiflorum , Salvia miltiorrhiza ,Salvia splendens , Tectona grandis ).. Based on 465 single-copy genes, MUSCLE was used for the sequence alignments and matrix construction. A maximum-likelihood phylogenetic tree was then constructed by IQtree2 (v20151210) with the ‘MFP+MERGE’ model. The divergence time of among 24 plants was predicted using BEAST2 (v2.5.4) (http://www.beast2.org/) with a strict molecular clock model, and time scales were calibrated by the divergence time of the fossil record of species from TimeTree (http://www.timetree.org). The results of OrtherFinder, CAFÉ 5 (https://github.com/hahnlab/CAFE) (Mendes, Vanderpool, Fulton, & Hahn, 2020) were used to analyze the expansion and contraction of gene families. Gene families were regarded as significantly expanded or contracted if the p-value was less than 0.05 in all species. Covariance analysis within genes were performed by jcvi (Tang et al., 2015). A gene pair was considered to have a covariance relationship while cscore was greater than 0.7. Homologous gene pairs were marked with strips and target genes with colorful trips.