with their estimated regression coefficients and associated p-values for each of the five phenotypes are listed in Supplemental Table 3. Of note is that for each of the 31 SNPs the estimated regression coefficient was in the same direction across all 5 phenotypes. Among the 31 SNPS, we identified 13 that were also significant at p < 10-3 for at least two additional phenotypes, suggesting they might be of relevance for the behavioral disinhibition construct. Results for these SNPs are given in Table 2. Only one SNP (rs804292, which maps to the 3’ UTR of NEIL2, a gene involved in DNA repair with no obvious relevance for substance abuse) had a p-value < 10-5 for more than one of the five phenotypes (Alcohol Dependence and Nicotine). Across all the markers and phenotypes, the lowest observed p value was 4.9 × 10-8 for rs1868152 for the Illicit Drug phenotype. This SNP is more than 30kb downstream of the CRYGS gene on 3q27.3. CRYGS codes for an eye lens protein that has been implicated in cataract formation but has not been previously linked to substance abuse or behavioral disinhibition. Only one of the gene assignments for the top 13 SNPs has any
