BackgroundRoses are important plants for human beings with pivotal economical and biological traits like continuous flowering, flower architecture, color and scent. Due to frequent hybridization and high genome heterozygosity, classification of roses and their relatives remains a big challenge.ResultsHere, to identify potential markers for phylogenetic reconstruction and to reveal the patterns of natural selection in roses, we generated sets of high quality and comprehensive reference transcriptomes for Rosa chinensis Old Blush' (OB) and R. wichuriana Basye's Thornless' (BT), two species exhibiting contrasted traits of high economical importance. The assembled reference transcriptomes showed transcripts N50 above 2000bp. Two roses shared about 10,073 transcripts (N50=2282bp), in which a set of 5959 transcripts was conserved within genera of Rosa. Further comparison with species in Rosaceae identified 4447 transcripts being common (Rosaceae-common) in Rosa, Malus, Prunus, Rubus, and Fragaria, while a pool of 164 transcripts being specific for roses (Rosa-specific). Among the Rosaceae-common transcripts, 409 transcripts showed a signature of positive selection and a clustered expression in different tissues. Interestingly, nine of these rapidly evolving genes were related to DNA damage repair and responses to environmental stimulus, a potential associated with genome confliction post hybridization. Coincident with this fast evolution pattern in rose genes, 24F-box and four TMV resistant proteins were significantly enriched in the Rosa-specific genes.ConclusionsWe expect that these Rosaceae-common and Rosa-specific transcripts should facilitate the phylogenetic analysis of Rosaceae plants as well as investigations of Rosa-specific biology. The data reported here could provide fundamental genomic tools and knowledge critical for understanding the biology and domestication of roses and for roses breeding.