Surface-enhanced Raman scattering (SERS) of genomic DNA was used to determine genetic relationships and species identification of nine plants from three subfamilies of Rosaceae. Genomic DNA was extracted, and the SERS spectra were obtained by using a nanosilver collosol at an excitation wavelength of 785 nm. Adenine and ribodesose were the active sites of genomic DNAs in the silver surface-enhanced Raman spectra. The strong peak at 714 cm(-1) was assigned to the stretching vibration of adenine, the strong peak at 1011cm(-1) contributed to the stretching vibration of the deoxyribose and the scissoring vibrations of cytosine, and the strong peak at 625 cm(-1) is the stretching vibration of glycosidic bond and the scissoring vibrations of guanine. The three-dimensional plot of the first, second, and third principal components showed that the nine species could be classified into three categories (three subfamilies), consistent with the traditional classification. The model of the hierarchical cluster combined with the principal component of the second derivative was more reasonable. The results of the cluster analysis showed that apricot (Prunus armeniaca L.) and cherry (Prunus seudocerasus Lindl.) were clustered into one category (Prunoideae); firethorn (Firethorn fortuneana Li.), loquat (Eriobotrya japonica Lindl.), apple (Malus pumila Mill.), and crabapple (Malus hallianna Koehne.) were clustered into a second category (Pomoideae); and potentilla (Potentilla fulgens Wall.), rose (Rosa chinensis Jacd.), and strawberry (Fragaria chiloensis Duchesne.) were clustered into a third category (Rosoideae). These classifications were in accordance with the traditional classification with a correction rate of clustering of 100%. The correct rate of species identification was 100%. These five main results indicate that the genetic relationship and species identification of nine Rosaceae species could be determined by using SERS spectra of their genomic DNAs.