Anemia is a unique side effect in Korean gastric cancer patients after S-1 monotherapy. We studied gastric cancer patients from a phase II trial of S-1 monotherapy with a 2-week treatment and 1-week rest schedule. Patients from a phase II trial of S-1 monotherapy with a 4-week treatment and 2-week rest were used as a reference group. The patients were categorized into two groups based on the degree of hemoglobin reduction per cycle of S-1: the mild reduction group (MRG DeltaHb/cycle < or =1.0) or severe reduction group (SRG DeltaHb/cycle >1.0). DeltaHb/cycle was calculated from maximum reduction of hemoglobin per one cycle of the treatment. Microarray-CGH was performed using a 17K cDNA microarray containing 15,723 unique genes. We selected genes with copy number variation defined as amplification (log2R/G >0.68) or deletion (log2R/G <-0.68), and a genetic aberration frequency difference of > or =30% between the MRG and the SRG. There were no differences in clinical factors, S-1 treatment-related factors (dose, dose intensity), toxicity, S-1 metabolism-related gene copy numbers (CYP2A6, DPD), or progression-free survival between the MRG and the SRG. Three genes were selected from microarray-CGH and logistic regression model: logit LN(Z) = (1.321) + (1.038 x PTX1) + (0.211 x MYO5A) + (0.516 x ZNF664). In the SRG, all 3 genes showed a trend of higher copy numbers than the MRG. There were no common anemia-related genes identified from different chemotherapy schedule of S-1 monotherapy. The logistics obtained from 3 genes predicted the hemoglobin reduction with an accuracy of 78%. The AUC was 0.744 for the final regression model. The combined copy number changes of the 3 genes can be developed into a biomarker in predicting S-1 treatment-related anemia.