The Promise of Genetics in Alcohol Use Disorders and the Problems of Phenotype, Polygenetic Architecture, Ancestry, and Comorbidity

With the advent of the sequencing of the human genome (1, 2), there was great promise that genetics would finally clarify the biological contributions to the “causes” of psychiatric disorders. The heritability of psychiatric disorders, including alcohol use disorder (AUD), had long been documented at between 34% and 77% (approximately 50% for AUD) based on twin and family studies (3–5), making their genetic underpinnings seemingly tractable. Three major waves of genetic studies over the past two decades have identified susceptibility loci for AUD (6–10). Most notably, genes related to ethanol metabolism were successfully identified first in candidate gene studies and later in genome-wide association studies (GWASs) (11–17). These findings represent important progress but explain only a small fraction of the observed heritability in alcohol phenotypes. Various factors have been proposed to account for this gap—commonly referred to as “missing heritability”—including small sample sizes (18–22). Yet, even the largest GWASs to date, with samples exceeding 900,000 individuals, accounted for just 4.2% of the variance in alcohol consumption and 6.6% in problematic use (23, 24). This suggests that sample size alone does not explain the gap.

Another challenge in the field has been the limitations of the “common disorder–common variant” hypothesis. It was discovered early on that unlike rare Mendelian disorders such as Huntington’s disease that are caused by rare, single, high-penetrance gene variants (25), common disorders like diabetes and mental health disorders including AUD do not arise from individual common variants of large effect. Instead, extensive reviews and meta-analyses of large genetic datasets have concluded that many psychiatric disorders are best characterized by a multifactorial etiology and a highly complex polygenic architecture (26). One important alternative framework is to embrace the presence of genetic heterogeneity (27), wherein different genetic pathways may underlie subtypes or dimensions of risk for AUD. Such heterogeneity may also arise across ancestry groups, where allele frequencies, linkage disequilibrium patterns, and ancestry-specific variants contribute to differences in genetic risk. Accounting for this heterogeneity, whether across phenotypes, disease progression, or ancestry, has already been shown to enhance gene discovery and improve the interpretability of genetic findings (28–30).

In this issue, Na et al. (31) adopt this strategy and use a quantitative measure of additive genetic risk for alcohol-related phenotypes, the polygenic risk score (PRS). PRSs can be constructed to capture the aggregate influence of common genetic variants and provide a quantitative index of liability for AUD. Although PRSs typically explain only a small proportion of variance in alcohol phenotypes (32), they are valuable tools for exploring how aggregate genetic risk manifests across different contexts, comorbidities, and populations. To illustrate, a recent study (33) demonstrated that PRS for alcohol dependence significantly predicted a greater likelihood of alcohol dependence diagnostic status and a more rapid rate of progression from drinking initiation to disorder. These applications show that PRSs can potentially capture heterogeneity, indicating not just who might develop AUD but also its timing and form. Drinking initiation was the only drinking milestone examined in that study, however, raising the question of whether similar relationships would be observed for other alcohol-related phenotypes.

Na et al. extend this strategy in a crucial way. Their study is the first to utilize large samples of people of African (AFR) and European (EUR) ancestry to examine the direct and interactive effects of genetic (biological), psychiatric, and environmental factors in predicting AUD criterion count, a measure of AUD severity. Rather than relying on alcohol consumption measures, they constructed PRSs from a problematic alcohol use GWAS (24), which aligns more closely with clinical severity and progression. In doing so, they embraced the heterogeneity perspective: focusing on a phenotype more likely to capture genetic variance relevant to disorder liability. To further strengthen prediction across ancestries, they applied a cross-population PRS method that combines ancestry-specific predictors. They found significant genetic associations with AUD criterion count in both ancestries, which explained ~1.3% of variance in the AFR sample and ~1.9% in the EUR sample. Although modest, these effects confirm that polygenic risk contributes to AUD severity across ancestries. Importantly, the associations remained significant even after accounting for psychiatric and environmental factors, highlighting the independent role of common genetic variants.

At the same time, a combination of education, early exposure to household substance use, income, and male sex explained 73.1% of the variance in AUD criterion count in the AFR sample, and 58.9% in the EUR sample. Among examined psychiatric disorders, posttraumatic stress disorder (PTSD) explained the most variance (10.0% in AFR and 9.4% in EUR), followed by anxiety disorders (3.4% in AFR and 6.2% in EUR) and major depressive disorder (1.3% in AFR and 2.1% in EUR). These findings highlight the enormous weight of trauma and comorbidity in shaping AUD severity. The authors also identified moderation effects: In individuals of European descent, higher education influenced the impact of problematic alcohol use PRS on AUD severity. This buffering effect, absent in participants of African descent and requiring cautious interpretation, suggests that social investments in educational retention may mitigate genetic vulnerability—although cross-sectional data cannot rule out reverse causation (i.e., AUD limiting educational attainment). Still, it offers a compelling illustration of how polygenic liability and social environment intertwine.

One psychiatric disorder not evaluated by Na et al. is antisocial personality disorder, long tied to an externalizing form of AUD. Descriptive typologies proposed by Babor (34) and Cloninger et al. (35–37) based on internalizing/externalizing psychopathology, age at onset, and severity distinguished between the more heritable antisocial subtype (type II) and the more environmentally driven, “milieu-limited” subtype (type I). Since Na et al. did not report on the influence of antisocial personality disorder on phenotypes, they may not have fully captured genetic variance associated with “antisocial alcoholism.” A Swedish family-based AUD study (38) reinforced this point, with a three-class solution. The three classes were type 1, male preponderant and externalizing (32%); type 2, minimal prior psychopathology (46%); and type 3, mixed-sex and internalizing (23%). The study showed that while all classes significantly aggregated in affected relative pairs, type 1 had the highest genetic risk for externalizing disorders and AUD whereas type 3 had genetic liability to internalizing disorders in relatives. The study also highlighted the power of family-based research and demonstrated that supposed genetic causes of AUD may arise from several disorders and dimensions, some of which may be unique to different racial populations.

Insufficient genetic diversity in study populations has been a persistent challenge in genetic studies, where most participants are of European ancestry (39, 40). The present study by Na et al. and other studies published by this group have identified important sources of both genetic and environmental variance across several ethnic groups, including African Americans. The genetic model fits both single variant and interactions; however, it was stronger in participants of European ancestry than in those of African ancestry, reflecting the Eurocentric bias of current GWASs and emphasizing the urgent need for large-scale, ancestry-diverse discovery efforts to advance both scientific accuracy and health equity. Finally, the heritability of AUD is well established, but its genetic architecture is complex, heterogeneous, and context dependent. PRSs reflect the common disease–common variant model, and their predictive utility depends on careful phenotype definition and consideration of disease progression and comorbidity. The small explanatory power of PRSs points to the limits of additive common variants and highlights the clear need for more comprehensive genetic models to better capture the biology of AUD and other psychiatric disorders.

The implications of the Na et al. study are broad. PRSs are not yet tools for clinicians, but they demonstrate that genetic liability is measurable and should be considered alongside family history, trauma exposure, and social determinants. For researchers, the study highlights the need to 1) expand GWASs in diverse populations, 2) refine phenotyping to capture AUD progression and heterogeneity, and 3) develop more comprehensive genetic models to reflect the complexity of AUD. For policymakers, a key lesson is that environments are modifiable: Improving educational access, reducing childhood exposure to substance use, and mitigating poverty-related stressors may buffer genetic risk at scale.