To establish a method for predicting the response to chemotherapy for osteosarcoma (OS), we performed expression profile analysis using cDNA microarray consisting of 23,040 genes. Hierarchical clustering based on the expression profiles of 19 biopsy samples of OS demonstrated two major clusters, one of which consisted exclusively of typical OS, i.e. conventional central OS in long bone of patients in the second decade. A set of genes was identified to characterize this subgroup, some of which have previously indicated some relation to carcinogenesis. Thirteen of the 19 patients were treated with an identical protocol of chemotherapy containing doxorubicin, cis-platinum and ifosfamide, and histological examination of resected specimens after operation classified 6 cases as responder and 7 as non-responder. A comparison of expression profiles of these two groups identified 60 genes whose expression levels were likely to be correlated with the response to chemotherapy (P<0.008). A drug response scoring (DRS) system was developed based on the expression levels of these genes, which proved to be applicable to predict the response to chemotherapy irrespective for the subclassification of OS. The reliability of the DRS system was further confirmed by testing additional 5 OS cases. These results indicated that scoring system based on gene-expression profiles might be useful to predict the response to chemotherapy for OS.