One of the most critical issues to be solved in regard to cancer chemotherapy is the establishment of ways to predict the efficacy of anti-cancer drugs for individual patients. To develop a prediction system based on expression of specific genes, we analyzed expression profiles of mononuclear cells from 18 chronic myeloid leukemia (CML) patients who were treated with the tyrosine kinase inhibitor STI571. cDNA microarrays representing 23 040 genes identified 79 genes that were expressed differentially between responders and non-responders to STI571. On the basis of the expression patterns of 15 or 30 of these genes among the patients, we developed a "Prediction Score" system that could clearly separate the responder group from the non-responder group. Verification of this system using four additional ("test") cases succeeded in predicting the response of each of those four patients to the drug. These results provide the first evidence that gene-expression profiles can predict sensitivity of CML cells to STI571, and may eventually lead to the achievement of "personalized therapy" for this disease.