prior probabilities of observed response and nonresponse in the absence of genotyping increased to posterior probabilities of 88% and 85%, respectively. For the remaining 
110 cases with predicted probability of response between 30% and 80%, posterior probabilities did not differ significantly from prior probabilities of response or nonresponse. 59% of observed nonresponders in the validation sample with 86% accuracy. Cross-validation of prediction model PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19672638 Validation of prediction model In the validation sample of 176 new completer cases treated with SSRI drugs, the distributions of clinical characteristics and genetic markers did not differ from the derivation sample. Seven of the top 10 SNPs that showed the strongest association with SSRI treatment response in the derivation sample were significantly BCTC web associated in the validation sample. In this validation cohort, the PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19675955 prior probabilities were 66% for response and 34% for nonresponse . The HAP-SNP model made predictions for 60% of these patients, with 84 predicted responders and 22 predicted nonresponders. The observed outcomes in these 106 predicted cases were 74 responders and 32 nonresponders. The overall accuracy of prediction was 85%. Sensitivity and specificity were 96% and 59%, respectively. This specificity result was not significantly lower than in the derivation sample. Among these 106 cases, the prior probabilities of observed response and nonresponse in the absence of genotyping increased to posterior probabilities of 85% and 86%, respectively. We found associations of SSRI response with 4 SNPs in TPH2. A previous small study found that three SNPs in TPH2, rs1843809 and rs1386492 of intron 5, and rs1487276 of intron 8, were associated with drug response after 12 weeks of SSRI treatment . However, there was no significant association between those three SNPs and SSRI response in the present study. Another study performed in a European population investigated nine SNPs in the TPH2 gene, and found two SNPs, rs10879346 and rs1487278, were significantly related to antidepressant response. Additionally, rs2171363 was significant in a Discussion The markers associated with response to SSRI drugs comprised ten SNPs from the TPH2, SLC6A4, GRIK2, and GAD1 genes and six haplotypes from the TPH2, SLC6A4, and GRIK2 genes. Thus, SSRI response was associated with polymorphisms in serotonin, glutamate, and GABA related genes. TPH2 showed the most significant association with SSRI response. TPH2 encodes the rate-limiting enzyme of brain serotonin production. Genetic Prediction of SSRI Response Chinese population. We imputed these three SNPs using genotype data, because they were not genotyped in our study. The imputed SNPs showed significant associations with SSRI response. These results from imputed data increase the possibility that the predictive markers suggested in our data will be replicable in other populations. Our previous studies indicated that two VNTRs in the 2nd intron and promoter of SLC6A4 are associated with SSRI response. In this study, we examined the two VNTRs and 12 SNPs in the SLC6A4 gene, and found both VNTRs and two SNPs were significantly associated with SSRI response. The two SNPs, rs2066713 and rs2020942, were located in intron 1 and intron 3, respectively. A previous study reported that these two SNPs have no association with SSRI response at 12 weeks in an ethnically mixed population, but another U.S. study reported that rs2066713 showed a trend towards association w