Substance dependence diagnoses (SDs) are important risk factors for suicidality. We investigated the associations of multiple SDs with different suicidality outcomes, testing how genetic background moderates these associations. The Yale-Penn cohort (N = 15,557) was recruited to investigate the genetics of SDs. The Army STARRS (Study to Assess Risk and Resilience in Servicemembers) cohort (N = 11,236) was recruited to evaluate mental health risk and resilience among Army personnel. We applied multivariate logistic regression to investigate the associations of SDs with suicidality and, in the Yale-Penn cohort, we used the structured linear mixed model (StructLMM) to study multivariate gene–environment interactions. In Yale-Penn, lifetime polysubstance dependence was strongly associated with lifetime suicidality: having five SDs showed an association with suicidality, from odds ratio (OR) = 6.77 (95% confidence interval, CI = 5.74–7.99) for suicidal ideation (SI) to OR = 3.61 (95% CI = 2.7–4.86) for suicide attempt (SA). In Army STARRS, having multiple substance use disorders for alcohol and/or drugs was associated with increased suicidality ranging from OR = 2.88 (95% CI = 2.6–3.19) for SI to OR = 3.92 (95% CI = 3.19–4.81) for SA. In Yale-Penn, we identified multivariate gene–environment interactions (Bayes factors, BF > 0) of SI with respect to a gene cluster on chromosome 16 (LCAT, p = 1.82 × 10–7; TSNAXIP1, p = 2.13 × 10−7; CENPT, p = 2.32 × 10−7; PARD6A, p = 5.57 × 10−7) for opioid dependence (BF = 12.2), cocaine dependence (BF = 12.1), nicotine dependence (BF = 9.2), and polysubstance dependence (BF = 2.1). Comorbidity of multiple SDs is a significant associated with suicidality and heritability of suicidality is partially moderated by multivariate gene interactions.