Purpose: To identify genetic biomarkers that might contribute to gemcitabine response and prediction of treatment outcome for pancreatic cancer by performing association of SNPs with patient overall survival (OS) of pancreatic patients treated with gemcitabine.Experimental Design: We used top 200 SNPs identified from previous GWAS to perform associations of these SNPs with OS using 400 patients (discovery samples)treated with gemcitabine using Cox proportional hazards models.We also performed SNP-gemcitabine interaction analysis to determine the effect of SNP on gemcitabine treatment response.All of these analyses were followed by imputation of+/-200 kb sides of regions surrounding top OS-associated SNPs.Additional 524 DNA samples were used to replicate the top candidate SNPs using TaqMan assay.Functional validation with siRNA knockdown for candidate genes was then performed.Results: Four top SNPs located in chromosome 1, 3, 9, and 20were identified to be associated with OS (P< 0.01) during the SNP-gemcitabine interaction analysis.We then imputed these four genomic regions followed by association of the imputed SNPs with our phenotype.Subsequently, we selected top 4 imputed SNPs plus those 4 genotyped SNPs to perform a replication study with additional 537 samples using TaqMan SNP genotyping assay followed by SNP-gemcitabine interaction analysis.The replication analysis showed that two imputed SNPs located in LOC285401 and CDH4 had trends in associations with OS during the treatment of gemcitabine.Functional validation demonstrated that knockdown of CDH4 significantly desensitized pancreatic cancer cells to gemcitabine cytotoxicity.Conclusion: We identified associations of SNPs with outcome during the gemcitabine treatment and the CDH4 gene might be involved in gemcitabine response in pancreatic cancer.Our genotype-phenotype association and functional study would enhance our understanding of the mechanisms of gemcitabine action and might help better predict gemcitabine response in pancreatic cancer.