Several genome-wide association (GWA) studies of alcohol dependence (AD) and alcohol-related phenotypes have been conducted; however, little is known about genetic variants influencing alcohol withdrawal symptoms (AWS). We conducted the first GWA study of AWS using 461 cases of AD with AWS and 408 controls in Caucasian population in the Collaborative Study on the Genetics of Alcoholism (COGA) sample. Logistic regression analysis of AWS as a binary trait, adjusted for age and sex, was performed using PLINK. We identified 51 SNPs associated with AWS with p<10 -4. The first best signal was rs770182 (p = 3.65 × 10 -6) at 5q21 near EFNA5 gene which was replicated in the Australian twin-family study of 273 families (p = 0.0172). Furthermore, three SNPs (rs10975990, rs10758821 and rs1407862) within KDM4C gene at 9p24.1 showed p<10 -4 (p = 7.15 × 10 -6, 2.79 × 10 -5 and 4.93 × 10 -5, respectively) in the COGA sample while one SNP rs12001158 within KDM4C with p = 1.97 × 10 -4 in the COGA sample was replicated in the family sample (p = 0.01). Haplotype analysis further supported the associations of single-marker analyses of KDM4C in the COGA sample. Moreover, two SNPs (rs2046593 and rs10497668) near FSIP2 at 2q32.1 with moderate associations with AWS in the COGA sample (p = 2.66 × 10 -4 and 9.48 × 10 -5, respectively) were replicated in the family sample (p = 0.0013 and 0.0162, respectively). In addition, several SNPs in GABRA1, GABRG1, and GABRG3 were associated with AWS (p<10 -2) in the COGA sample. In conclusion, we identified several loci associated with AWS. These findings offer the potential for new insights into the pathogenesis of AD and AWS.