Genetic substructure in Latin American individuals reveals novel associations, mechanistic insights, and variable polygenic risk score transferability for alcohol traits

Genetic substructure in Latin American individuals reveals novel associations, mechanistic insights, and variable polygenic risk score transferability for alcohol traits | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Article Genetic substructure in Latin American individuals reveals novel associations, mechanistic insights, and variable polygenic risk score transferability for alcohol traits Janitza Montalvo-Ortiz, José Martínez-Magaña, Diego Andrade Brito, and 63 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8789707/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Genome-wide association studies (GWAS) have substantially advanced our understanding of the genetic architecture underlying alcohol consumption. However, Latin American populations represent only ~ 1.8% of participants in current GWAS. Here, we present the largest GWAS meta-analysis of alcohol consumption in Latin American populations to date, analyzing 465,516 individuals through the Latin American Genomics Consortium (LAGC). We identified 14 independent loci, including 13 previously known associations and one novel locus in WRN . Multi-omic integrative network analysis revealed two functional modules: synaptic signaling pathways and inflammatory response mechanisms, extending beyond alcohol metabolism genes. Polygenic risk score (PRS) transferability varied substantially across Latin American subgroups. This study-derived PRS outperformed European-derived scores in South Americans and Puerto Ricans, while European PRS performed better in Mexicans and Cubans. Unsupervised genetic clustering confirmed that PRS performance depends on ancestral composition rather than geographic labels. These findings expand our understanding of the genetics of alcohol consumption in Latin Americans by identifying novel associations and demonstrating significant genetic heterogeneity within Latin American populations. Results underscore that population-specific approaches are essential to ensure broadly applicable genomic medicine. Health sciences/Medical research/Genetics research Biological sciences/Genetics/Medical genetics Alcohol consumption genome-wide association study meta-analysis Latin America genetic risk score Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Alcohol consumption patterns vary substantially across global populations, with excessive consumption contributing to over 60 chronic diseases and increased mortality from road traffic and other accidents 1 – 3 , making it a leading cause of disability and death worldwide. Latin American countries exhibit distinct alcohol consumption patterns that contribute significantly to regional alcohol-related mortality burdens, though comprehensive epidemiological data are limited to some countries 4 – 6 . Notably, the prevalence of alcohol-related diseases, such as alcohol-associated hepatitis and cirrhosis 7 , is rising across Latin America 8 , even in countries with relatively low per capita alcohol consumption 9 , 10 , underscoring the growing negative impact of alcohol consumption in this region. Alcohol consumption traits result from complex interactions between environmental and genetic factors Genome-wide association studies (GWAS) have substantially advanced our understanding of the genetic architecture underlying alcohol consumption and related disorders 11 – 18 . However, these studies have disproportionately focused on individuals of European ancestry 19 , 20 , with Latin American populations representing less than 2% of participants according to the GWAS Diversity Monitor 21 . This limited representation is problematic due to the low transferability of genetic findings from European populations to Latin American populations, attributed to differences in linkage disequilibrium (LD) patterns, allele frequencies, and environmental exposures, among other factors 22 including possible differences in underlying genetic architecture. To date, a few large-scale GWAS of alcohol consumption have included Latin American individuals (sample sizes: 14,112–286,026 Latin American individuals), identifying loci such as GCKR , chromosome 4 alcohol metabolizing enzyme genes, and DDX31 12,16,23 . However, the heritability explained by these variants in Latin American populations is lower than in European populations, highlighting the critical need for broader representation to better understand the genetic architecture of alcohol traits in these populations. Latin America encompasses South America, Central America, Mexico, and the Caribbean regions characterized by predominant Latin-derived languages and a remarkable cultural and environmental landscape 24 . The genetic landscape of Latin America reflects multiple waves of historical admixture between different ancestral groups, creating unique genetic compositions across countries 25 – 30 . This admixture significantly impacts health-relevant genetic variation 31 . Additionally, substantial migration from Latin America to North America, particularly to the United States of America (USA) where Latin American populations represent a rapidly growing demographic 32 , has influenced alcohol consumption patterns in these communities 33 – 36 . Previous alcohol-related GWAS in Latin American populations have frequently relied on USA-recruited cohorts, potentially limiting representation of the full clinical, environmental, geographic, and ancestral diversity across these populations. Here, we conducted a comprehensive multi-trait meta-analysis of alcohol consumption GWAS of 465,516 individuals of Latin American ancestry from North, Central, and South America (Fig. 1). Our objectives were to: 1) identify genetic loci associated with alcohol consumption traits in Latin American populations, 2) characterize their biological functionality and potential causal variants, 3) investigate genetic correlations with other health-related phenotypes, and 4) evaluate PRS transferability across these populations. This work represents a collaborative effort by the Latin American Genomics Consortium (LAGC, https://www.latinamericangenomicsconsortium.org/ ) , established in 2019 and affiliated with the Psychiatric Genomics Consortium (PGC) to advance psychiatric genomics research in Latin American populations 24 . Diversifying GWAS across Latin American populations is essential for discovering novel disease associations and ensuring equitable access to the benefits of genomic medicine in these communities 20 , 37 – 39 . Results Geographically diverse data collection across Latin American populations We assembled a comprehensive dataset of Latin American individuals across multiple cohorts from different geographical regions. The primary meta-analysis included 465,516 individuals of Latin American ancestry from nine cohorts ( Supplementary Table 1 ), classified using both geographical origin and genetic ancestry criteria (Fig. 2 a ). Eight country- or region-specific cohorts contributed to the meta-analysis: Mexico City Prospective Study (MCPS) 40 , 41 , Mexican Genomic Database for Addiction Research (MxGDAR) 42 , Brazilian High-Risk Cohort (BHRCS) 43 , Boston Puerto Rican Health Study (BPRHS) , 44 , VA Million Veteran Program (MVP) 44 , All of Us Research Program (AoU) 45 , Hispanic Community Health Study/Study of Latinos (HCHS/SOL) 46 , and Spit for Science (S4S) 47 (Fig. 2 a ). Additionally, we included multicountry data from 23andMe customers of Latin American descent 16 . For supplementary analyses (e.g. PRS, complementary ADH locus analysis, and ancestry-specific allelic frequency) and to strengthen representation across Latin America, we incorporated five additional country-specific cohorts (Fig. 2 b ), El Banco por Salud (USA) 48 , São Paulo Epidemiologic Sleep Study (EPISONO, Brazil) 49 , Baependi Heart Study (BHS, Brazil) 50 , Metabolic Syndrome Cohort (Mexico) 51 , and newly genotyped data from the MxGDAR and COVID-19 Puerto Rico (COVID19-PR, Puerto Rico) study. We used several definitions of alcohol consumption across cohorts, including drinks per week, AUDIT-C 52 scores, and individual items of the AUDIT-C test, detailed in Supplementary Table 1 . We performed principal components analysis of cohorts from the main meta-analysis with available individual genetic data, including HCHS/SOL (USA), MxGDAR-Fz1 (Mexico), BHRCS (Brazil), and BPRHS (USA). Our analysis confirmed the substantial genetic diversity characteristic of Latin American populations, revealing complex ancestral patterns across these cohorts (Fig. 2 c ). Genome-wide associations of alcohol consumption The fixed-effect sample size-weighted meta-analysis across cohorts using METAL 53 identified 1,203 genome-wide significant (GWS) variants (Fig. 3 a , Supplementary Table 2 ). Using a sliding window (250 kb, LD r 2 = 0.2), we defined 14 independent variants ( Supplementary Table 3 ), thirteen of which were previously associated with alcohol consumption in other populations. Quality control metrics indicated negligible inflation, with the lambda (λ) of genomic control (GC) equal to 1.11 (Fig. 3 b ) and LDSC intercept 54 was 1.06 (SE 0.01). The associated variants mapped to established alcohol-associated genes including GCKR *rs1260326 (chr2: 27508073:T:C), CADM2 *rs2167046 (chr3:85537656:G:A), KLB *rs11940694 (chr4:39413373:A:G), EPHA3 *rs73137382 (chr3:89412213:C:G), TSPAN5 *rs184703805 (chr4:98516868:G:A), ADH1B* rs1229984 (chr4:99318162:T:C), DDX31 *rs10736856 (chr9:132615400:C:T), NAA25 *rs11066132 (chr12:112030402:C:T), along with four intergenic variants: rs60478839 (chr4:98972194:G:A), rs1789896 (chr4:99335827:G:A), rs148926542 (chr4:100386039:A:G), rs10065698 (chr5:12080851:C:A). Regional association plots are provided in Supplementary Fig. 1 . We also replicated a previously identified lead variant associated with the interaction between alcohol consumption and high-density lipoprotein in TRIB1AL *rs2954021 (chr8:125469835:A:G) 55 . The strongest association was observed for the missense variant ADH1B* rs1229984 (P = 1.12×10 − 203 ) (as for previous EUR GWAS of similar traits). Additionally, we detected rs671 (chr12:111803962:G:A, ALDH2*2 ), a variant predominantly found in Asian populations 56 , showing significant association in our Latin American sample (P value = 2.09×10 − 32 ), and present in 23andMe and MCPS cohorts. We identified one novel locus, at rs34431249 (chr8:31151900:T:A), an intronic variant on WRN , which encodes a RecQ helicase essential for maintaining genomic stability 57 . The novel lead variant has showed associations with regional brain volumes 58 . Our study identified more than twice the number of independent variants of previous alcohol consumption GWAS that included Latin American populations (Fig. 3 c ). SNP-based heritability (h 2 g ) was significant at 2.13% (s.e. 0.002) on the observed scale (Fig. 3 d ), comparable to previous estimates in Latin American populations (e.g., 3.2% for drinks per week) 16 but higher than AUDIT-C based estimates 12 . Statistical fine-mapping using SuSiE 59 identified 14 potential causal variants with posterior inclusion probability (PIP) > 0.9 ( Supplementary Table 4 ), including KLB *rs11940694 (PIP = 0.94), TSPAN5 *rs184703805 (PIP = 1.0), and ADH1B* rs1229984 (PIP = 1.0). We also tested the association of ADH locus variants with drinks per week in an independent Mexican cohort (Metabolic Syndrome Cohort, n = 5,095) 51 , but observed no GWS associations ( Supplementary Table 5 ), likely due to limited statistical power. We used several definitions of alcohol consumption across cohorts, including drinks per week, AUDIT-C scores, and individual items of the AUDIT-C test. Given this phenotypic heterogeneity across cohorts, we performed a complementary multi-trait meta-analysis across the cohorts with significant heritability, MCPS, MVP, and 23andMe using MTAG 60 . This analysis identified 856 GWS variants ( Supplementary Fig. 2a , Supplementary Table 6 ) and using a sliding window (250 kb, LD r 2 = 0.2), we identified six independent variants: GCKR *rs1260326 (chr2: 27508073:T:C), CADM2 *rs2167046 (chr3:85537656:G:A), KLB *rs13135439 (chr4:39403531:T:G), ADH1B *rs1229984 (chr4:99318162:T:C), ADH1C *rs1662048 (chr4:99350875:T:C), and TRIB1AL *rs2980868 (chr8:125475993:C:T), and one intergenic variant (rs10065698, chr5:12080851:C:A) (Supplementary Table 7 ). Quality metrics remained robust (λ GC = 1.11, LDSC intercept = 1.02 ± 0.02) ( Supplementary Fig. 2b ), with significant SNP-based heritability of h 2 g = 2.10% (s.e. = 0.02) ( Supplementary Fig. 2c ). The strongest association was again with ADH1B* rs1229984 (P = 1.28×10 − 196 ). Fourteen GWS variants, located within the same loci identified in the fixed-effect meta-analysis (mapping to CADM2 , MTTP , and CTNND2 genes), were unique to the multi-trait analysis ( Supplementary Fig. 2d , Supplementary Table 8 ), demonstrating the complementary value of this approach. Ancestry-specific allelic frequencies reveal population-specific genetic architecture Given the admixed genetic composition of Latin American populations, comprising primarily European (EUR), African (AFR), and Admixed American (AMR) ancestries, we investigated whether minor allelic frequencies (MAF) of alcohol-associated variants differed across these ancestral backgrounds. An analysis of independent and causal variants using 1000 Genomes 61 and GnomAD reference 62 data revealed that six independent variants exhibited higher allele frequencies in EUR compared to other ancestral populations ( Supplementary Fig. 3A ). The frequency of the ADH1B* rs1229984 (AMR-MAF = 0.06 and EUR-MAF = 0.04) was highest in individuals of AMR ancestry, distinguishing it from other variants in the ADH locus that were more frequent in EUR, such as rs1789896 (AMR-MAF = 0.36 and EUR-MAF = 0.50). Highlighting the potential effects of previous reports of selection pressures on shaping the allelic frequencies of the variants inside this locus 63 , 64 . We further evaluated the allele frequencies of the independent variants across each analyzed cohort. While we generally observed consistent frequencies across cohorts, there were notable exceptions ( Supplementary Fig. 3B ). For instance, the KLB *rs11940694 and intergenic rs2954021 variants showed substantial lower frequency in the Mexican cohorts (MCPS and MxGDAR), of about 20%. In contrast, the intergenic variant rs148926542 was observed at higher frequency in the BPRHS cohort of Puerto Rican descent. To examine ancestry-specific effects more precisely, we performed local ancestry inference in seven cohorts (HCHS/SOL, MxGDAR, BHRCS, BPRHS, EPISONO, BHS, and COVID19-PR) and calculated the ancestry-specific allelic frequencies for 8,840,280 high-quality variants (imputation R 2 > 0.9) across the genome (Fig. 4 a ). Genome-wide analysis revealed a higher proportion of low allele frequency variants (MAF < 0.05) in AMR genomic segments, with this pattern diminishing as MAF increased. In contrast, GWS alcohol-associated variants identified by the fixed-effect meta-analysis showed enrichment in both EUR and AMR genomic segments (Fig. 4 b ). The frequency of the ALDH2*2 variant (rs671), known for its crucial role in alcohol metabolism 56 and pleiotropy across health outcomes 65 , 66 , often is high in East Asian populations 67 . We observed variable MAF’s across several Latin American cohorts included in several phases of this study, ranging from 0.0004 (BPRHS, Puerto Ricans living in the Boston) to 0.0054 (EPISONO, Brazilian cohort) (Fig. 4 c ), similar to that found the AMR individuals of the gnomAD database (0.0004) 62 . We also identified the variant in Mexican cohorts (MxGDAR) and individuals of Mexican descent living in the USA (El Banco). Local ancestry analysis in MCPS revealed that ALDH2*2 was exclusively present in individuals carrying AMR ancestry segments at this genomic location, highlighting the AMR origin of this variant in Latin American populations. Further analysis of the MAFs of the six independent variants identified in the fixed effect meta-analysis GWAS ( GCKR *rs1260326, EPHA3 *rs73137382, DDX31 *rs10736856, WRN *rs34431249, intergenic variants rs1789896 and rs10065698) demonstrated that they were consistently higher in EUR than AMR ancestry segments ( Supplementary Table 9 ), indicating that local ancestral background may shape the allelic frequencies of these alcohol-associated loci in Latin American populations. Multi-omic integration reveals functional architecture of alcohol consumption variants Variant annotation and genomic distribution Functional annotation of GWS variants using SNPnexus 68 revealed that 860 variants mapped to 45 genes, while 343 variants were intergenic ( Supplementary Fig. 4 ). CADM2 and MTTP each harbored over 100 associated variants (representing one independent locus), being the most GWS variant-dense loci. Among genic variants, 91.7% were in intronic positions, consistent with potential regulatory mechanisms underlying alcohol consumption genetics, given that introns may shape the gene regulatory landscape 69 , 70 . Additionally, we identify 10 protein-coding variants, GCKR *rs1260326 (chr2: 27508073:T:C), ADH1B *rs1229984 (chr4:99318162:T:C), ADH1B *rs698 (chr4:99339632:T:C), ADH1B *rs1693425 (chr4:99344955:C:T), ADH1B *rs1789915 (chr4:99345214:A:G), ADH7 *rs971074 (chr4:99420704:C:T), MTTP *rs3816873 (chr4:99583507:T:C), MTTP *rs113557405 (chr4:99591692:T:C), BRAP *rs3782886 (chr12:111672685:T:C), and ALDH2 *rs671 (chr12:111803962:G:A). Gene-based association analysis identified key functional genes MAGMA 71 gene-based association analysis using AMR individuals of the 1000 Genomes reference panel, to allow ancestry-matched LD estimation, identified 17 significantly associated genes: EIF2B4 , GCKR , CADM2 , RFC1 , KLB , TSPAN5 , METAP1 , ADH6 , ADH1B , ADH1C , ADH7 , C4orf17 , TRMT10A , MTTP , LOC285556 , DNAJB14 , and TMEM254 ( Supplementary Table 10 ). This gene set encodes established alcohol metabolism enzymes ( ADH gene cluster) and other genes involved in diverse biological processes. We performed multi-tissue and multi-cell-type chromatin-interaction analyses to capture shared regulatory effects and uncover convergent biological mechanisms linked to the associated loci. We applied H-MAGMA 72 analysis across 28 tissues and cell types, identifying 322 significant associations representing 35 unique genes ( Supplementary Fig. 5 , Supplementary Table 11 ). Nine genes showed consistent chromatin interaction associations across all analyzed tissues: GCKR , ADH1B , MTTP , TSPAN5 , C4orf17 , AC083902.1 , ADH1C , AC083902.2 , and ABT1P1 , suggesting broad regulatory importance in alcohol consumption pathways. Brain-specific regulatory chromatin interactions To identify brain-relevant regulatory mechanisms associated with our GWAS findings, we performed chromatin interaction mapping to overlap alcohol-associated variants with high-confidence regulatory chromatin interactions (HCRCI) in fetal and adult cortex tissues 73 . We evaluated fetal and adult cortex tissue to have a better understanding of conserved mechanisms across brain cortex development. Following Bonferroni correction for multiple testing, we identified 16 significant SNP-gene pairs mapping to two genes: SLC4A1AP (lowest P = 4.47x10 − 12 , with 32 Hi-C loops in adult and 15 Hi-C loops in fetal) and DNAJB14 (lowest P = 3.95x10 − 27 , with 13 Hi-C loops in adult and 19 Hi-C loops in fetal) ( Supplementary Table 12 ). The prioritization of additional genes highlights the complementary use of HCRCI to identify potential novel targets associated with alcohol consumption across cortex development. Transcriptome-wide association studies reveal tissue-specific gene expression effects We performed transcriptome-wide association studies (eTWAS) using S-PrediXcan 74 across 49 tissues to prioritize genes whose predicted expression levels are associated with our GWAS of alcohol consumption, giving insights into the functional roles of the associated variants in the regulation of gene expression. Single-tissue analysis identified 14 genes associated with alcohol consumption in at least one tissue, LAMTOR3 , FNDC4 , KLB , IFT172 , MTTP , TRMT10A , WDR19 , KRTCAP3 , METAP1 , SMTN , ADH7 , ADH1B , ADH1C , and RP11-766F14.2 ( Supplementary Table 13 ). We replicated the well-known associations seen in other ancestry groups in the alcohol metabolism enzymes. Integrative multi-tissue eTWAS using S-MultiXScan 75 identified 29 genes after Bonferroni correction, including established alcohol metabolism genes ( ADH1C , ADH1B , ADH7) , among others: RP11-766F14.2 , MTTP , LAMTOR3 , TRIM34 , SMTN , GCKR , METAP1 , TPRG1L , FNDC4 , GPR146 , KLB , TUT1 , RP11-484P15.1 , SIAH2 , RP11-789C1.2 , VWA7 , IFT172 , TRMT10A , RFC1 , DDIT4L , JPH3 , FUT2 , EIF2B4 , NRBP1 , ATF2 , and UCK1 ( Supplementary Table 14 ). To evaluate these findings with population-appropriate LD patterns, we performed complementary eTWAS using FUSION 76 with AMR individuals of the 1000 Genomes reference panel. Single-tissue analysis identified 14 genes after Bonferroni correction: RFC1 , DAPP1 , NRBP1 , KLB , MTTP , WDR19 , METAP1 , CADM2 , ADH1B , GPN1 , AC021148.1 , AC074117.1 , ZNF512 , and ADH1C ( Supplementary Table 15 ). Multi-tissue integration using sparse canonical correlation analysis and the Cauchy association test (ACAT) 77 identified 14 genes after Bonferroni correction: RFC1 , DAPP1 , NRBP1 , SNX17 , KLB , MTTP , WDR19 , METAP1 , CADM2 , ADH1B , GPN1 , AC074117.1 , ZNF512 , and ADH1C ( Supplementary Table 16 ). A cross-method comparison revealed 21 genes that were prioritized across single-tissue analyses ( Supplementary Fig. 6 ), while multi-tissue approaches identified seven genes with robust support across both SMultiXcan and ACAT methods: ADH1B , ADH1C , KLB , METAP1 , MTTP , NRBP1 , and RFC1 (Fig. 5 a ). A splicing transcriptome-wide association study (sTWAS) using S-PrediXcan 74 identified alternative splicing isoforms associated with alcohol consumption. Single-tissue analysis detected 33 splicing isoforms across six genes ( Supplementary Table 17 ): SNX17, RFC1, TSPAN5, ADH5, ADH1B, and ADH1C (Fig. 5 b ). Multi-tissue sTWAS identified 29 splicing isoforms after Bonferroni correction ( Supplementary Table 18 ), spanning 22 genes including alcohol metabolism enzymes ( ADH1B, ADH1C, ADH5 ) as well as PKP4, RRP8, SNX17, TOM1L2, NDUFA6-DT, GRB14, PACSIN2 , IRAK1BP1 , MAST4 , SDHAP4 , MRPS33 , PHIP , TMEM255B , TSPAN5 , ADAMTS9 , TPCN2 , SCYL2 , PCNAP1 , and TMEM254 . Proteome-wide association study (PWAS) reveals plasma proteins associated with alcohol consumption Using multi-ancestry trained models 78 , 79 , we used S-PrediXcan to identify plasma proteins associated with alcohol consumption. European American-trained models identified five significant proteins: ADH1C, ADH1A, ADH7, CADM2, and KLB ( Supplementary Table 19 ). African American-trained models identified two proteins: ADH1C and CADM2. When we compared the PWAS results with the Bonferroni significant eTWAS results, we identified convergent direction of effects between gene expression and protein abundance, with ADH1C and ADH7 showing negative z-scores, but inverse effects for KLB (positive in protein and negative in gene expression). No significant associations were detected using Latin American-trained models, partly due to limited statistical power (sample size, n = 301). Single-cell polygenic analysis reveals tissue-specific cellular mechanisms To characterize the single-cell polygenic architecture of alcohol consumption, we used scPagwas 80 to integrate our GWAS summary statistics with single-cell RNA sequencing data across brain regions (amygdala 81 , striatum 82 , cerebellum 83 , hypothalamus 83 , hippocampus 83 , and cortex 83 ) and peripheral tissues (adipose tissue 84 , arteries 85 , and liver 86 ). Several tissues (striatum, cerebellar vermis, adipose tissue, and arteries) showed no significant associations after background correction ( Supplementary Table 20 ). After background correction, significant cell-type associations were observed exclusively in the amygdala and liver. In the amygdala, genetic liability for alcohol consumption was significantly associated with both inhibitory neurons (P = 1.21×10 − 8 ) and excitatory neurons (P = 2.01×10 − 13 ) (Fig. 5 c, Supplementary Table 20 ). Gene ontology (GO) enrichment analysis of the alcohol-correlated genes (Pearson correlation coefficient > 0.1) in amygdala revealed significant enrichment for synaptic processes, including modulation of chemical synaptic transmission (adjusted P = 3.20×10 − 43 ), regulation of trans−synaptic signaling (adjusted P = 3.20×10 − 43 ) and synapse organization (adjusted P = 2.43×10 − 31 ) ( Supplementary Fig. 8, Supplementary Table 21 ). In liver tissue, genetic liability for alcohol consumption was associated with multiple cell types when analyzing common variants (MAF > 0.01): cholangiocytes (P = 1.81×10 − 9 ), endothelial cells (P = 1.86×10 − 8 ), fibroblasts (P = 2.62×10 − 12 ), and hepatocytes (P = 9.19×10 − 15 ) (Fig. 5 d , Supplementary Table 20) . These associations were enriched for metabolic pathways, including organic acid catabolic process (Adj. P = 1.19×10 − 13 ), carboxylic acid catabolic process (Adj. P = 1.19×10 − 13 ), and small molecule catabolic process (adjusted P = 3.66×10 − 12 ). When including low-frequency variants (MAF > 0.001), the liver cell type association shifted dramatically to immune cell populations: conventional dendritic cells (P = 6.97×10 − 13 ), macrophages (P = 8.56×10 − 9 ), monocytes (P = 1.29×10 − 13 ), neutrophils (P = 1.12×10 − 3 ), plasma cells (P = 8.81×10 − 3 ), and plasmacytoid dendritic cells (P = 2.90×10 − 11 ) ( Supplementary Fig. 7 ). This shift was accompanied by enrichment in immune-related pathways, including cytoplasmic translation (adjusted P = 4.80×10 − 100 ), cellular respiration (adjusted P = 1.56×10 − 32 ), and antigen processing and presentation of peptide antigen via MHC class II (adjusted P = 9.87×10 − 16 ), immune response−regulating signaling pathway (adjusted P = 2.67×10 − 7 e-07), and leukocyte-mediated immunity (adjusted P = 3.54×10 − 6 ). Cross-tissue analysis reveals shared and tissue-specific mechanisms Comparative analysis between the amygdala and liver identified MAPK1 as consistently correlated with alcohol consumption across both tissues ( Supplementary Fig. 9a ). Tissue-specific top correlated genes included RBFOX3 , PTPRR , CELF4 , and CACNA1B in amygdala, while liver showed distinct patterns depending on the MAF threshold: PIK3CD , KLRF1 , FTX , and CLASP1 for common variants, and RPS24 , RPL15 , RPL13 , EEF1A1 for analysis including low-frequency variants. Pathway analysis of shared genes between tissues revealed modulation of synaptic transmission as the top-enriched pathway ( Supplementary Fig. 9b, Supplementary Table 22 ), suggesting common signaling mechanisms underlying alcohol consumption effects across brain and peripheral tissues. Integrative Multi-omic Analysis We evaluated multiple gene prioritization methods to our GWAS findings to identify those that has the higher potential to be functional, then we evaluated genes that may have convergent effects across all the different prioritizing methods; suggesting conserved genes related with the genetic liability of alcohol consumption. Integrating the results of our multi-omic analyses identified 24 genes that were consistently prioritized by at least three independent methods (Fig. 5 d ). Among these, we replicated several genes previously associated with alcohol consumption in other populations, including ADH1C , ADH1B , ADH4 , ADH6 , ADH5 , ADH7 , TSPAN5 , MTTP , RFC1 , METAP1 , RAP1GDS1 , KLB , CADM2 , SNX17 , GCKR , TRMT10A , LAMTOR3 , WDR19 , NRBP1 , IFT172 , and C4orf17 . We also identified potentially novel candidate genes not previously implicated in alcohol consumption traits, including DAPP1 , FNDC4 , and SMTN . Given that the methods we used to prioritized genes of our GWAS have a high heterogeneity in the omics information used to trained them, we further investigate conserved functional relationships among the prioritized genes, we employed two recently developed network-based tools: GRIN 87 and MENTOR 88 . GRIN was used to assess the network connectivity of genes within a 33-layer multiplex network, retaining 31 highly connected genes of the 96 initial genes ( Supplementary Fig. 11, Supplementary Table 23 ). Notably, while GRIN retained most members of the alcohol and aldehyde dehydrogenase families ( ADH1A , ADH1B , ADH1C , ADH5 , ADH7 , and ALDH2 ), two genes ( ADH4 and ADH6 ) were not retained. We then applied MENTOR to cluster the 31 GRIN-retained genes into functional modules based on network embeddings (Fig. 5 f , Supplementary Table 24 , Supplementary Fig. 10 ). This analysis resolved two distinct modules: Module C1 comprised 14 genes ( ALDH2 , EMCN , SLC39A8 , MAPKAPK5 , MAPK1 , ACAD10 , HECTD4 , CADM2 , CTNND2 , TRAFD1 , NRBP1 , WDR19 , KRTCAP3 , and GTF3C2-AS2 ), while module C2 included 17 genes ( TMEM254 , RFC1 , ADH5 , METAP1 , DNAJB14 , ZNF512 , WRN , ATF2 , GPN1 , PPM1G , SNX17 , LAMTOR3 , H2AFZ , ADH7 , ADH1B , ADH1C , and ADH1A ). Module C2 captured the core alcohol-metabolizing enzyme genes, along with genes from loci newly implicated in this study. Functional enrichment of module C1 revealed genes involved in ERK MAPK signaling ( MAPK1 , MAPKAPK5 ) and synaptic signaling ( CADM2 and CTNND2 ). Interestingly, MAPK1 was uniquely prioritized by single-cell analyses across multiple tissues, but not by other multi-omic methods. In contrast, module C2 was enriched for genes implicated in inflammatory signaling ( ATF2 , LAMTOR3 , and WRN ), including those from novel loci identified in this study. Genetic correlations (rg) We assessed genetic correlations among alcohol consumption phenotypes across the largest contributing cohorts: 23andMe, MCPS, and MVP. We identified significant positive correlations between them ( Supplementary Fig. 12 ). To further evaluate the genetic overlap between alcohol consumption and other complex traits, we analyzed genetic correlations with 64 GWAS that included Latin American populations from the GWAS catalog 89 . Nominally significant genetic correlations were observed between alcohol consumption and several traits, including body mass index (BMI, rg = -0.08, s.e. = 0.02), high-density lipoprotein cholesterol (HDL, rg = 0.20, s.e. = 0.07), number of cups of coffee consumed per day (rg = 0.21, s.e = 0.08), and triglyceride levels (rg = -0.18, s.e. = 0.07). However, none of these correlations remained statistically significant after Bonferroni correction for multiple testing ( Supplementary Table 25 ). Alcohol consumption PRS performance across Latin American subgroups We evaluated the association between PRS and alcohol consumption - measured as drinks per week (DrinksWk) - in the HCHS/SOL cohort. To avoid overfitting, we constructed PRSs using summary statistics from the current meta-analysis, excluding HCHS/SOL. We also generated ancestry-specific PRSs based on previously published GWAS of individuals of African (AFR, n = 8,078) and European (EUR, n = 666,978) ancestry 16 . We employed PR-CS 90 for single-ancestry PRS construction and PRS-CSx 91 for multi-ancestry modeling. We first assessed the performance of PRSs across geographically defined subgroups in HCHS/SOL (Fig. 6 a ). Using PRS-CS, both the PRS constructed from our current study (PRS-CS - This Study = 8.47×10 − 5 ) and the PRS derived from the EUR GWAS (PRS-EUR = 2.77×10 − 6 ) were significantly associated with DrinksWk in the full HCHS/SOL sample ( Supplementary Table 26 ). The variance explained was modest: 0.21% for the PRS from this study and 0.30% for the EUR-derived PRS ( Supplementary Fig. 13a ). Additionally, we tested if adding cohorts even if with small sample sizes (n < 10000) increase the variance explained by the PRS. The PRS generated using only the largest cohorts (23andMe, MCPS, and MVP; which included the higher percentage of the samples included in this study) explained less variance (PRS-CS-Biobanks, 0.08%) ( Supplementary Fig. 13a, Supplementary Table 26 ) that the once build with all cohorts, suggesting that sample diversity enhances predictive accuracy. Similar findings were observed with PRS-CSx ( Supplementary Table 27 ). To examine heterogeneity across Latin American subpopulations, we stratified individuals by self-reported geographic origin: Central America (n = 754), Cuba (n = 1411), the Dominican Republic (n = 535), Puerto Rico (n = 955), Mexico (n = 2063), and South America (n = 490). We observed substantial variability in PRS performance across subgroups (Fig. 6 b, Supplementary Table 26 ). The PRS from this study was significantly associated with DrinkWk in individuals of South American (P = 0.0187) and Central American (P = 0.0378) descent, while the EUR-derived PRS was not (P > 0.17 and 0.54, respectively). Conversely, in individuals of Mexican (P = 0.0131) and Cuban (P = 7.00×10 − 4 ) descent, the EUR-derived PRS showed stronger associations than the study-derived PRS. In Puerto Rican individuals, both PRSs were significantly associated with DrinksWk (P = 0.0030 vs. PRSCs-EUR, P = 0.0140), but the study-derived PRS explained more variance (0.60% PRS in this study and 0.49% of the PRSCs-EUR). Similar results were obtained using PRS-CSx ( Supplementary Fig. 14, Supplementary Table 28 ). Except for Dominicans, for whom we observed a negative association, all other PRS showed positive betas, although the wide confidence intervals indicate substantial heterogeneity in the estimates. We further explored genetic substructure and applied K-means clustering to genetic principal components (PC1-PC4) in HCHS/SOL, identifying five clusters with optimal clustering performance based on the Silhouette score 92 ( Supplementary Fig. 15 ). In K-means, the silhouette score measures clustering quality, with values near 1 indicating well-separated clusters and values near 0 indicating overlapping clusters. This approach was applied to Latin American individuals from the MVP cohort in a previous study 93 . Significant associations with DrinkWk were detected in Cluster 1 (P = 0.0106 for this study PRS; P = 3.19×10 − 4 for EUR PRS) and Cluster 5 (P = 0.0025 and P = 0.0249, respectively) (Fig. 6 d ). In Cluster 5, the PRS from this study explained more variance than the EUR-derived PRS ( Supplementary Fig. 13a ). PRS-CSx also identified a significant association in Cluster 4 ( Supplementary Fig. 16; Supplementary Table 28 ). These clusters showed partial overlap with geographic origin: Cluster 1 was enriched for Cuban individuals, Cluster 3 for Mexican individuals, and Cluster 5 for Puerto Rican individuals (Fig. 6 d ). Admixture 94 analysis revealed that Clusters 1 and 5 had higher proportions of EUR ancestry (Fig. 6 e ). We extended the analysis to three independent cohorts: MXGDAR-Freeze2 (newly genotyped), El Banco por Salud (USA), and EPISONO (Brazil). In the Mexican-descent cohorts (MXGDAR-Freeze2 and El Banco por Salud), PRSs were not significantly associated with drinks per occasion ( Supplementary Table 28 ). However, in EPISONO, the PRS from this study was significantly associated with the alcohol use frequency, whereas the EUR-derived PRS was not ( Supplementary Table 28 ). Applying the K-means clustering approach to EPISONO identified two genetic clusters. Within these clusters, all three PRSs were significantly associated with alcohol use but the PRS constructed from the full meta-analysis explained the highest proportion of variance ( Supplementary Table 28 ). Discussion We conducted a GWAS meta-analysis of alcohol consumption traits in 465,516 Latin American populations, who are underrepresented in genomic research 21 . Our study helps to address this disparity by reporting the most extensive GWAS to date. Our findings substantially enhance understanding of the genetics of alcohol consumption in Latin American populations through integration of diverse functional approaches, including gene expression, splicing isoforms, protein levels, network analysis, and single-cell gene expression data. We explored genetic correlations and compared the predictive ability of PRS trained in Latin American populations against those from European populations, including novel comparisons across different geographical groups within the Latin American region. We estimated SNP-based heritability at 2.1%, comparable to a previous estimate of 3.2% for drinks per week 16 but higher than estimates for alcohol consumption using AUDIT-C scores 12 . AUDIT-C scores represent items that quantify the drinking frequency, typical consumption, and potential binge-drinking episodes, which provides a more comprehensive composite score than measures based solely on drinks per week. The modestly lower heritability compared to other studies including Latin American populations could be attributable to a higher genetic heterogeneity 54 as we included cohorts from ancestral backgrounds that are not well represented in LD reference panels (for example, Brazil, with less than 10 individuals included in HGDP and 1000 Genomes reference panel). This limited matching between the reference panels and target populations 54 , 96 can violate LD score regression assumptions, and is further complicated by population admixture effects 97 . These methodological challenges are particularly pronounced in Latin American populations, where currently available reference panels inadequately capture genetic diversity and current heritability estimation methods struggle with complex admixture patterns. We replicated genetic associations with alcohol consumption at several well-established loci. The strongest association was observed with ADH1B*rs1229984 , which encodes an enzyme with enhanced ethanol oxidation capacity 98 . This variant has been extensively linked to multiple alcohol-related traits 99 – 104 and replicated across various Latin American populations, including Mexican and Amerindigenous groups 105 . We also replicated an association with GCKR*rs1260326 , a variant involved in alcohol-related traits and several metabolic traits 17 , 99 , 106 , 107 , including type 2 diabetes, fasting insulin, and total cholesterol concentrations 99 , 108 – 111 . GCKR variants may play a critical role in modulating the effects of alcohol consumption on glucose and lipid metabolism. We identified associations in the CADM2 gene, previously linked to alcohol, smoking, 112 , and lifetime cannabis use 113 . CADM2 encodes a neural cell adhesion molecule and has been associated with a broad range of mental and metabolic phenotypes 113 – 123 . Additionally, KLB* rs11940694 has been associated with alcohol consumption across multiple studies, with the A alelle associated with reduced drinking 11 , 17 , 99 , 102 , 124 . We found that ALDH2 *2 (rs671) may be specific to Amerindigenous genetic ancestry in Latin American populations. This variant has long been associated with alcohol-related traits in East Asian populations 56 , 125 , 126 . The presence of the ALDH2*2 genetic variant in Latin American populations raises important questions about its potential origin. This variant is found at high frequency in East Asian populations, although its evolutionary origin remains unknown 127 , 128 . Previous studies in Amerindigenous groups from the USA reported very low frequency of this allele 129 . However, ALDH2 has undergone selection pressures in Andean populations 130 , for high altitude adaptation, even though no specific results were reported for the ALDH2*2 genetic variant. Additionally, an association between ALDH2*2 and alcohol use disorder has been reported in a Brazilian population, where the frequency of carriers of this variant was observed at approximately 4% 131 . When we analyzed ancestry-specific allelic frequencies using local ancestry inference, we observed higher frequencies of most GWS signals in individuals with EUR and AMR genomic segments, suggesting enrichment of specific alleles in these ancestral backgrounds. We highlight the need for more comprehensive investigations of the geographic and population-specific distribution of genetic variants across Latin America. Such studies will be essential for understanding the evolutionary history of this allele in the region and may inform translational and public health approaches for individuals carrying this variant 132 . Multi-omic analysis identified and prioritized potential functional gene targets associated with alcohol consumption. We replicated seven well-known alcohol-associated genes - ADH1C , ADH1B , RFC1 , MTTP , KLB , METAP1 , and CADM2 16 -with evidence across several methods supporting their relevance to alcohol consumption. Our findings align with previous GWAS results. This convergence further supports that associated loci in Latin American populations are shared with other ancestral groups, highlighting the robustness of these loci in relation to alcohol consumption. Alternative splicing contributes substantially to alcohol-related traits, accounting for approximately 30% of genetic risk and 2.3% of heritability 133 , 134 . Expanding on this, we identified splicing isoforms in genes such as SNX17 and ADH6 , both previously linked to alcohol use 11 , 135 . Notably, SNX17 may be involved in excitatory synapse loss 136 , underscoring the importance of studying the potential functional effects of the diversity of protein isoforms promoted by splicing events 137 . Our blood protein analysis identified four proteins, ADH1C, KLB, CADM2, and ADH7, associated with alcohol consumption genetic liability. CADM2 has been associated with multiple phenotypes in humans and rodent models 121 , including risky behaviors and metabolism-related traits such as body mass index, and mouse models suggest that reducing its expression could reverse metabolic syndrome-associated traits 115 , pointing to a pleiotropic effect of CADM2 in behavioral and metabolic traits. Previous studies have demonstrated that genetic liability for alcohol consumption is enriched in brain regions 12 . We extended these findings by exploring the effects of genetic liability for alcohol consumption at the single-cell level. For brain cell types, the implication of inhibitory neurons in the amygdala is consistent with prior single-nucleus analyses in alcohol withdrawal 138 . Our findings in the amygdala support the hypothesis that the genetic liability for alcohol consumption may be associated with synaptic transmission 139 . Finally, we identified pleiotropic cellular effects of genes linking alcohol consumption with cancer and immune function, such as MAPK1 , which was consistently associated with alcohol consumption genetic liability across several tissues. MAPK1 , a key gene in the mitogen-activated protein kinase pathway, mediates detrimental effects of alcohol 140 and influences binge-drinking behavior 141 . Moreover, in the liver, we identified associations with epithelial cell changes previously observed in single-cell analyses of patients with alcohol-related liver disease 142 . Together, our integrated multi-omic analysis revealed convergent evidence across gene expression, splicing, protein levels, and single-cell analyses, identifying both established alcohol metabolism pathways and novel mechanisms linking genetic liability to tissue-specific cellular processes in Latin American populations. Using a novel systems biology approach, we explored the effects of genes prioritized across multiple methods to identify potential conserved mechanisms influenced by alcohol consumption associated genes. Using GRIN 87 we identified 31 highly interconnected genes from 122 different gene prioritization methods and MENTOR 88 clustered these 31 GRIN-retained genes into two functional modules, highlighting synaptic transmission and inflammation as conserved mechanisms. As expected, genes clustered based on biological function rather than similar gene prioritization methods, according to their MENTOR-derived network embeddings. The MAPK/ERK signaling pathway genes MAPK1 and MAPKAPK5 were identified in the MENTOR module C1. This pathway has previously been associated with alcohol addiction in animal models 140 , 141 . CADM2 and CTNND2 , which encode scaffolding proteins involved in synaptic plasticity 121 , 143 , also clustered in module C1. Three genes involved in inflammatory signaling pathways clustered in module C2: WRN (which binds to the NF-kB complex) 144 , the transcription factor ATF2 145 , and LAMTOR3 , a member of the Regulator complex 146 . Moreover, genes from our novel loci, such as WRN , clustered with known alcohol-metabolizing genes, suggesting that they may contribute to similar functional effects within these gene networks. Our network-based analysis revealed functional stratification underlying alcohol consumption genetics that operate through multiple biological pathways, extending beyond traditional alcohol metabolism genes to encompass neural plasticity, cellular signaling, and immune response mechanisms across diverse tissues. We found that the PRS derived from our Latin American discovery sample provided better predictions in target samples from similar populations than those derived from European samples from previous large-scale GWAS. However, we observed significant variability in PRS performance across Latin American individuals of different geographical origins, suggesting that transferability varies across Latin American populations and underscoring the need for better characterization and representation of these diverse groups. This has previously been observed for PRS derived for kidney traits 147 . In addition to the heterogeneity observed in geographically stratified analyses, we also identified heterogeneity using a clustering model without prior labels. This phenomenon has been observed previously, for example in cardiovascular disease, where high heterogeneity in PRS transferability was noted across different genetic clusters of individuals of Latin American origin in the MVP 93 . We also observed greater transferability of the PRS derived from our study in individuals of Latin American descent from South America and Puerto Rico. When applying the clustering method without incorporating a prior geographical label, transferability of the PRS derived from our study was higher among individuals of Latin American descent with a greater proportion of European ancestry. These findings underscore two key points: first, the advantage of conducting well-powered GWAS in Latin American populations, allowing for reaching similar performance of the EUR PRS. Second, it highlights a critical gap in PRS transferability among different Latin American subgroups, consequently exacerbating health disparities within Latin America. One hypothesis that could explain the heterogeneity in transferability of the PRS is the lack of inclusion of individuals of Amerindigenous descent in large-scale GWAS, which could adversely impact the PRS estimation. Another possibility is that inadequate representation of these populations in reference panels for imputation or genotyping fails to capture genetic variants enriched in individuals of Latin American descent. While genetic diversity likely contributes to this heterogeneity, non-genetic factors, such as sociocultural differences, may also play a role 148 . These findings demonstrate that effective polygenic prediction in Latin American populations requires population-specific approaches that account for genetic diversity and ancestry composition, highlighting the need for tailored genomic medicine strategies. We performed the largest meta-analysis of alcohol consumption in Latin American populations to date, substantially increasing representation particularly of individuals from Latin American countries. We functionally characterized the genetic architecture of alcohol consumption in Latin America using multiple approaches, including novel single-cell and network-based methods. Importantly, we demonstrated greater PRS transferability using our Latin American GWAS compared to European-derived polygenic scores, while revealing performance variation across different Latin American subgroups. These findings demonstrate that genetic diversity within Latin American populations requires population-specific genomic approaches to ensure equitable prevision medicine implementation. Limitations Although our study advances understanding of the genetic architecture of alcohol consumption in Latin American populations, including individuals from both Latin American countries and the US, several limitations should be noted. Because most individuals in our meta-analysis derive from MCPS and 23andMe, diverse populations remain underrepresented, including those of low socioeconomic status and from additional geographical regions. Differences in ascertainment strategies and phenotypic definitions across cohorts may have introduced bias in our results. We addressed heterogeneity in alcohol consumption phenotypes across cohorts through MTAG analysis. Additionally, the lack of reference panels and functional genomic data that adequately capture Latin American genetic diversity limits post-GWAS functional characterization of identified associations. Further, the absence of large-scale multi-omic studies similar to GTEx, or brain single-cell initiatives, in Latin American populations limits the interpretation and generalizability of our functional results 149 – 151 . We partially addressed this limitation by using an ancestry-matched LD reference panel. Nevertheless, additional generation of Latin American–specific datasets is needed to identify potential population-specific results. Future studies incorporating comprehensive environmental assessments across Latin American populations could enable identification of gene-environment interactions, potentially revealing novel insights into complex trait etiology, as demonstrated in other populations 152 , 153 . Conclusion We conducted a large-scale GWAS meta-analysis for alcohol consumption in Latin American populations, advancing knowledge in three critical areas. First, through comprehensive multi-omic approaches and novel integrative network analyses, we identified and functionally characterized genes associated with alcohol consumption, revealing dual biological pathways involving synaptic signaling and inflammatory responses that extend beyond alcohol metabolism. Second, we demonstrated that PRS transferability varies substantially across Latin American subgroups, with scores outperforming European-derived predictions in some populations. Third, our findings suggest that genetic signals previously thought to be exclusive to a single ancestral group, such as the association at the ALDH2 locus, may also be present in other populations. These findings establish that Latin American populations exhibit significant genetic heterogeneity requiring population-specific approaches, highlighting the critical need for enhanced representation of diverse Latin American groups to ensure equitable precision medicine implementation. Material and methods Ethics All studies in this meta-analysis received appropriate ethical approval and followed relevant regulations for human subjects research. Informed consent was obtained from all participants or their legal guardians where applicable. The Mexico City Prospective Study (MCPS) was approved by scientific and ethics committees within the Mexican National Council of Science and Technology (0595 P-M), the Mexican Ministry of Health, and the Central Oxford Research Ethics Committee (C99.260). The Million Veteran Program (MVP) study was approved by the U.S. Department of Veterans Affairs’ central institutional review board (IRB). All 23andMe research participants provided informed consent and participated voluntarily in online research under a protocol approved by the external AAHRPP-accredited IRB, Ethical & Independent (E&I) Review Services. As of 2022, E&I Review Services became part of Salus IRB ( https://www.versiticlinicaltrials.org/salusirb ). For the Hispanic Community Health Study/Study of Latinos (HCHS/SOL), institutional review boards at each field center approved the study, and all participants provided written informed consent 154 . The Mexican Genomic Database for Addiction Research (MxDGAR) study obtained written informed consent or assent from all participants, with protocols reviewed and approved by the Research Ethics Committee of the Instituto Nacional de Psiquiatría (No. CEI/C/083/2015) and the Instituto Nacional de Medicina Genómica (No. 01/2017/I) in Mexico. The Brazilian High-Risk Cohort Study (BHRCS) was approved by the ethics committee of the University of São Paulo [IORG0004884/National Council of Health Registry number (CONEP): 15.457/Project IRB registration number: 1132/08]. Written consent was obtained from the participants' parents and from participants who could read, write, and understand the written consent form. The Boston Puerto Rican Health Study (BPRHS) was approved by the IRB at Tufts Medical Center and Northeastern University, with all participants providing written informed consent. The El Banco por Salud project was reviewed and approved by the University of Arizona Human Subjects Protection Program (#1703274963). All participants provided written informed consent from bilingual research staff to enroll in El Banco por Salud in their preferred language. The EPISONO study was approved by the Ethics Committee for Research at the Universidade Federal de São Paulo (UNIFESP) (CEP #0593/06) and registered with ClinicalTrials.gov (NCT00596713), with informed consent obtained from all participants. The Spit for Science study protocols were approved by the IRB of Virginia Commonwealth University. The metabolic syndrome cohort project was approved by the Ethics Committee of the Instituto Nacional de Ciencias Médicas y Nutrición, and all participants signed an informed consent form. The Baependi Heart Study (BHS) protocol conformed to the tenets of the Declaration of Helsinki and was approved by the Ethics Committee of the Hospital das Clinicas, University of São Paulo, Brazil (approval number 0494/10). The COVID-19 Puerto Rico (COVID19-PR) cohort was approved by the IRB from the University of Puerto Rico and Yale University, and consent was provided by all the participants. Study design and phenotype This study examined multiple alcohol consumption phenotypes, including individual items from the Alcohol Use Disorders Identification Test (AUDIT-C) 52 , the sum of the three AUDIT-C consumption items, and the number of drinks consumed per week. The primary meta-analysis study included 465,516 individuals of Latin American descent from nine independent cohorts recruited from Latin American countries (country-specific cohorts) and one multi-country cohort (23andMe). We conducted a complementary association analysis of genetic variants in the ADH locus in an independent sample of 5,053 from a Mexican cohort 51 . Polygenic risk score (PRS) analyses were performed across multiple cohorts, HCHS/SOL, additional newly genotyped individuals from MxGDAR-freeze 2 (n = 624), and two additional independent country-specific cohorts [El Banco por Salud 48 (n = 680) and EPISONO 49 (n = 565)]. For ancestry-specific allelic frequency analyses, we included two additional independent cohorts: BHS (n = 307) 50 and COVID19-PR (n = 242). Sample sizes were determined based on available participants from existing cohorts; no statistical power calculation was used to predetermine sample size requirements. Study cohorts in the primary meta-analysis Country-specific cohorts Mexico City Prospective Study (MCPS, n = 140,444) The Mexico City Prospective Study (MCPS) is a longitudinal study including up to 150,000 individuals residing in two municipalities in Mexico City (Coyoacán and Iztapalapa) who were aged at least 35 years at recruitment between 1998 and 2004 40,41 . We assessed two questions regarding alcohol consumption: (1) “How often did you have a drink containing alcohol in the past year?” (frequency), and (2) “On a typical occasion, how many cups or glasses of alcoholic beverages would the participant normally drink?” (quantity). We mapped these responses to AUDIT-C scoring criteria as follows. Quantity was scored as: None/I do not drink = 0, 1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 drinks = 4). Frequency was scored as: Never consumed alcohol = 0, never in the last 12 months but previously consumed/1–5 times per year/6–11 times per year = 1, approximately once per month/2–3 times per month = 2, 1–2 times per week = 3, 3–4 times per week/daily = 4. We also classified individuals as never versus ever drinkers and identified heavy drinkers according to National Institute on Alcohol Abuse and Alcoholism (NIAAA) criteria (men: ≥5 drinks on any day or ≥ 15 per week; women: ≥4 drinks on any day or ≥ 8 per week). Samples were genotyped using an Illumina Global Screening Array (GSA), and quality control (QC) steps as previously described 41 . Genotypes were imputed using the TOPMed imputation server. Post-imputation filtering excluded rare (minor allele frequency [MAF] < 0.05) and low-quality (INFO < 0.8) variants. For genetic association analysis, we used logistic regression models for ever drinker and heavy drinking phenotypes and linear regression models for frequency and quantity, implemented in SUGEN 155 and adjusted for age, sex, and seven principal components (PCs) of genetic ancestry. We performed Multi-Trait Analysis of GWAS (MTAG) 60 for the four phenotypes assessed in the MCPS cohort using the AMR-classified individuals from the 1000 Genomes reference panel 156 . Mexican Genomic Database for Addiction Research (MxGDAR, n = 3,403) We analyzed a subsample of the Mexican Genomic Database for Addiction Research (MxGDAR) 42 , a population-based cohort with national representativity across Mexico, designed to identify environmental and genetic factors contributing to substance use disorders, including alcohol consumption. We accessed the question: “When you drink alcoholic beverages, how many cups do you drink on each occasion?” and mapped responses to AUDIT-C scoring: None/I do not drink = 0, 1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 = 4). Genotyping was conducted using the Illumina PsychArray. For directly genotyped variants, we excluded variants with Hardy-Weinberg equilibrium (HWE) P ≤ 1x10 − 10 , and retained variants with MAF > 0.01. Genotypes were imputed using the TOPMed Imputation Server 157 . Imputed variants were excluded if INFO < 0.8 and MAF < 0.05. We performed association analysis using a mixed linear model for drinks per week using GCTA 158 , adjusting for age, sex, genetic relationship matrix (GRM) to account for cryptic relatedness, and five PCs of genetic ancestry. Brazilian High-Risk Cohort Study (BHRCS, n = 1,626) The Brazilian High-Risk Cohort for Psychiatric Disorders (BHRC) is a large, community-based longitudinal study - initiated in 2010 and now in its fourth follow-up - tracking 2,511 Brazilian youths (ages at the baseline) from São Paulo and Porto Alegre, enriched for mental conditions, with longitudinal deep clinical, cognitive, environmental, neuroimaging, and genome-wide genetic assessments to delineate risk factors and developmental trajectories of common mental disorders 43 . We assessed alcohol consumption in the past 12 months using the question: “How many drinks/alcoholic drinks do you have on a typical day when you are drinking?”. Responses were mapped to AUDIT-C scoring: None/I do not drink = 0, 1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 drinks = 4. Genotyping was conducted using the Illumina Global Screening Array. For directly genotyped variants, we excluded variants with a HWE P ≤ 1x10 − 10 and retained variants with MAF > 0.01. Genotypes were imputed using the TOPMed Imputation Server. Imputed variants were excluded if INFO < 0.8 and MAF < 0.05. We performed association analysis using mixed linear models for drinks per week implemented in GCTA 158 , adjusting for age, sex, GRM to account for cryptic relatedness, and five PCs of genetic ancestry. Boston Puerto Rican Health Study (BPRHS, n = 1,386) The Boston Puerto Rican Health Study (BPRHS) is a longitudinal study investigating the role of psychosocial stress on allostatic load and health outcomes in self-identified Puerto Ricans aged 45–75 years residing in Boston 159 . We accessed lifetime alcohol consumption using AUDIT-C-based questions for current drinkers: (“On average, on the days that you drink alcohol, how many drinks do you have a day?”) and former drinkers (“On average, on the days that you drank alcohol, how many drinks did you have a day?”). Responses were mapped to AUDIT-C scoring: None/I do not drink = 0, 1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 = 4. Genotyping was conducted using Affymetrix Axiom Genome-Wide LAT Array. For directly genotyped variants, we excluded variants with HWE P ≤ 1x10 − 10 and retained variants with MAF > 0.01. Genotypes were imputed using the TOPMed Imputation Server. Imputed variants were excluded if INFO < 0.8 and MAF < 0.05. We performed mixed linear models for drinks per week implemented in GCTA 158 , adjusting for age, sex, GRM to account for cryptic relatedness, and five PCs of genetic ancestry. Million Veteran Program (MVP, n = 31,877) The VA Million Veteran Program (MVP) is an observational cohort and mega-biobank of the Department of Veterans Affairs 44 aimed at understanding how health is affected by genetic characteristics, behaviors, and environmental factors. We evaluated alcohol consumption using the maximum AUDIT-C score 52 , which ranges from 0–12 and is calculated from responses to three questions: (1) “How often did you have a drink containing alcohol in the past year?” (never = 0, monthly or less = 1, 2–4 times per month = 2, 2–3 times per week = 3, ≥ 4 times a week = 4), (2) “How many drinks did you have on a typical day when you were drinking in the past year?” (1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 = 4); (3) “How often did you have six or more drinks on one occasion in the past year? (never = 0, less than monthly = 1, monthly = 2, weekly = 3, daily or almost daily = 4). We used summary statistics from a previous MVP GWAS that evaluated maximum AUDIT-C scores 23 . Genotyping was performed using a custom Affymetrix Axiom Biobank Array, with analyses using MVP Release 3 data. Quality control was performed by the MVP Genomics working group prior to imputation. Samples with excessive heterozygosity, a missing call rate > 2.5%, or variants with a low call rate or deviation from the expected allele frequency were removed. Genotypes were phased and imputed with EAGLE 160 and Minimac 161 with the 1000 Genomes Project phase 3 reference panel. We excluded one individual randomly from each pair of related individuals (kinship coefficient = 0.0884). Variants were excluded based on MAF < 0.005, genotype call rate ≤ 0.95, and HWE P ≤ 1x10 − 6 . Variants with INFO scores < 0.3 were removed using SNPTEST 162 . The MVP cohort includes individuals from various ancestral backgrounds, with Latin American ancestry assignment performed using the Harmonized Ancestry and Race/Ethnicity (HARE) method 163 . Genetic association analysis was performed using linear regression models in PLINK, version 1.9 164 , with covariates including age at maximum AUDIT-C score, sex, and the first ten. All of Us Research Program (AoU, n = 10,200) The All of Us Research Program (AoU) is a USA initiative to enroll a diverse group of at least 1 million participants to accelerate biomedical research and improve health. All of Us began enrollment in 2018 and currently enrolls participants aged 18 or older from more than 340 recruitment sites across the country 45 . We assessed alcohol consumption using the AUDIT-C question: “On a typical day, when you drink, how many drinks do you have?”. Responses were mapped to AUDIT-C scoring: None/I do not drink = 0, 1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 = 4. This study included all individuals who self-identified as Latino or Hispanic in release 6 of the All of Us cohort with available whole-genome sequencing data 165 . We excluded related individuals (kinship coefficient = 0.1) as established by All of Us , estimated using PCrelate 166 implemented in Hail 167 . We excluded variants with HWE P ≤ 1x10 − 10 and retained variants with MAF > 0.01. Genetic association analysis was performed using linear models implemented in Hail adjusted with age, sex at birth, and ten PCs of genetic ancestry. Hispanic Community Health Study/Study of Latinos (HCHS/SOL, n = 6,076) The Hispanic Community Health Study/Study of Latinos (HCHS/SOL) is a large, multicenter, population-based probabilistic sample of Hispanic/Latino individuals aged 18–74 recruited in New York City, Chicago, Miami, and San Diego. Participants’ heritage was self-reported as Puerto Rican, Cuban, Dominican, Mexican, Central American, or South American 46 , 168 , 169 . We used drinks per week as the measure of alcohol consumption 170 . Participants were genotyped using a custom Illumina array, including HumanOmni2.5-8v1-1 array content plus approximately 150,000 investigator-chosen SNPs 171 . We used freeze 3 data, where genotypes were imputed using IMPUTE2 172 with the 1000 Genomes Phase 3 reference panel. We excluded SNPs with R 2 < 0.3 and performed association analysis using mixed linear models for drinks per week implemented in GCTA 158 , with GRM to account for cryptic relatedness, and adjustment for age, sex, and five PCs of genetic ancestry. Spit for Science (S4S, n = 950) Participants for the S4S study were included from an ongoing longitudinal cohort study of college students at a large, urban, mid-Atlantic public university investigating the complex interplay between genetic, environmental, and developmental factors contributing to substance use and related behaviors. This study was approved by the university’s review board and all participants provided informed consent. For a detailed review of study methods see (Dick et al., 2014) 47 . Study data were collected and managed using REDCap electronic data capture tools 173 , 174 . Alcohol consumption was assessed using the question: “How many drinks did you have on a typical day when you were drinking in the past year?”. Responses were mapped to AUDIT-C scoring: None/I do not drink = 0, 1–2 drinks = 0, 3–4 drinks = 1, 5–6 drinks = 2, 7–9 drinks = 3, ≥ 10 = 4. Genetic ancestry was determined using ten ancestry PCs and then theminimum Mahalanobis distance was calculated and ancestry was assigned to the closest reference population (e.g., Admixed Americas) 175 . Genotyping was performed at Rutgers University Cell and DNA Repository using the Affymetrix BioBank array. Imputation was conducted using SHAPEIT2 176 and IMPUTE2 172 using the 1000 Genomes Phase 3 reference panel. Post-imputation filtering removed rare (MAF < 0.05) and low-quality (INFO < 0.8) variants. Association analysis was performed using mixed linear models implemented in GCTA-MLMA 158 with a leave-one-chromosome-out protocol, including GRM to account for cryptic relatedness and adjustment for age, sex, and five PCs of genetic ancestry. Initial GWAS was conducted in three batches for the recruitment cohorts (n = 445, 207, 298), followed by inverse-variance-weighted fixed-effect meta-analysis using METASOFT 177 . Multi-country cohorts 23andMe (n = 279,007) Participants were drawn from 23andMe, Inc.'s consumer genetics database. We used previously published summary statistics for drinks per week 16 . Ancestry assignment was performed using 23andMe’s local ancestry method, which splits genomic data into short windows of approximately 300 SNPs. Haplotypes are classified using multiple reference populations derived from the Human Genome Diversity Project, HapMap 178 , 1000 Genomes, and 23andMe customers who reported having four grandparents from the same country. A hidden Markov model assigns probabilities for each reference population, with classification thresholds defined by 23andMe. Samples were genotyped using 23andMe genotyping platforms, with genotypes imputed against the 1000 Genomes Project Phase 1 reference using SHAPEIT2 179 . Genetic association tests were performed using linear regression adjusted for age, sex, and the top five PCs. Meta-analysis of GWAS We conducted genome-wide association meta-analyses using the METAL software 53 , applying a sample size-weighted scheme. For each variant, a combined z-statistic and corresponding P were calculated on a weighted sum of individual study z-scores, with weights proportional to the square root of the effective sample size. Prior to meta-analysis, we harmonized the summary statistics across studies using GWASLab 180 , which included alignment to the reference genome, annotating with dbSNP rsID, and genome build liftover where necessary. We excluded insertions or deletions annotated as D/I, or if the variants sample size was less than 15% of the total meta-analysis sample size. After filtering, a total of 18,690,427 SNPs were included in the meta-analysis. We assessed genomic inflation using the genomic control lambda (λ GC) and linkage disequilibrium score regression (LDSC) intercept 54 , to account for residual population stratification and polygenicity. Given the heterogeneity in alcohol consumption phenotype definitions across cohorts, we also conducted a multi-trait analysis using MTAG 60 , which jointly analyzes genetically correlated traits while accounting for sample overlap. This analysis combined summary statistics from 23andMe, MCPS, and MVP. For MTAG, we restricted SNPs with MAF > 0.01, non-ambiguous strands, and sample sizes ≥ two-thirds of the 90th percentile of the sample size distribution, resulting in 6,107,401 SNPs retained for analysis. Complementary analysis of the ADH gene locus We further examined the ADH gene locus; we performed a complementary analysis in an independent cohort of 5,095 individuals from the metabolic syndrome cohort 51 . Genotyping was performed using the Illumina Global Diversity Array, and genotype quality control included exclusion of variants with HWE P ≤ 1x10 − 10 and retention of variants with MAF > 0.01. Genotypes were imputed using the TOPMed Imputation Server. We extracted GWS variants at the ADH locus identified in our meta-analysis and tested their association with the number of drinks consumed per week. Association testing was conducted using linear regression adjusted for age, sex, and the first ten PCs of genetic ancestry, implemented in R. Associations were considered statistically significant at a false discovery rate (FDR) – adjusted P < 0.05. Identification of independent variants To identify independent signals within associated regions, we performed LD-based clumping using a 250 kb window and an LD threshold of r 2 > 0.1, using individuals of AMR ancestry from the 1000 Genomes Project as the LD reference. Loci containing independent variants located within 1Mb of each other were merged into a single genomic region. Due to the extensive long-range LD in the ADH gene cluster on chromosome 4, all variants within this region were merged into a single locus. For loci with multiple independent signals, we conducted approximate conditional analyses using GCTA-COJO 181 . Variants that remained GWS (P < 5x10 − 8 ) after conditioning were included in interactive conditional analyses, sequentially adding the next most significant variant until no further independent associations were detected. To assess novelty, we cross-referenced all independent lead variants with the NHGRI-EBI GWAS Catalog 89 and screened relevant literature on alcohol consumption and alcohol use disorder. Fine-mapping We performed statistical fine-mapping of each GWS locus using the Sum of Single Effects regression model (SuSiE) model 59 , a Bayesian variable selection approach that decomposes association signals into sparse, single-effect components. Fine-mapping was conducted using GWAS summary statistics and LD information derived from AMR individuals in the 1000 Genomes Project. Variants were considered putatively causal if their posterior inclusion probability (PIP) exceeded 0.90. Ancestry-specific allele frequency estimation We estimated ancestry-specific frequencies using local ancestry inference across multiple cohorts, including HCHS/SOL, MxGDAR, BHRCS, BPRHS, EPISONO, and two additional independent cohorts, the Baependi Heart Study (n = 307) 50 and the COVID-19-PR cohorts (n = 242). Variants were filtered for MAF > 0.001 and imputation INFO > 0.9. Genotypes were phased using SHAPEIT5 182 , and local ancestry was called using RFMix2 183 , with a combined reference panel of AMR, AFR, and EUR ancestry from the 1000 Genomes Project and Human Genome Diversity Project 95 (total n = 2,306) RFMix2 was run with parameters -n 5 and -e 1 to account for sample size imbalance and admixture in the reference. Ancestry-specific allelic frequencies were calculated using Tractor 184 . For the MCPS cohort, allele frequencies were retrieved from the Regeneron Genetics Center variant browser ( https://rgc-mcps.regeneron.com/ ). SNP-based heritability (h 2 ) We estimated SNP-based heritability (h2)-the proportion of phenotypic variance explained by common variants-using LDSC 54 , restricted to HapMap3 SNPs 178 . LD reference panels were derived from AMR individuals in the 1000 Genomes Project Phase 3. Observed scale h 2 was estimated separately for the METAL meta-analysis (1,206,107 SNPs) and MTAG analysis (531,758 SNPs). For sensitivity analysis, we also tested alternative LD reference panels, including EUR individuals from 100 Genomes and the Slim Initiative in Genomic Medicine for the Americas (SIGMA) cohort, using covariate-adjusted LDSC (cov-LDSC) 185 , which yielded comparable heritability estimates. SNP annotation All GWS variants were annotated using SNPnexus 68 , employing dbSNP rsIDs to map each variant to genomic coordinates, associated genes (RefSeq), and genic elements including coding regions, introns, and untranslated regions (5′-UTR, 3′-UTR), as well as noncoding regions. Gene-based association analyses Gene-level association testing was conducted using MAGMA 71 with default parameters. We applied the default mean association model using the summary statistics of the meta-analyzed data using METAL and for annotation we used the protein-coding genes regions from NCBI released with the software. LD information was derived from AMR individuals in the 1000 Genomes Project. Bonferroni correction for the number of tested genes (P < 0.05/18,124 genes = 2.76x10 − 6 ) was used to define statistical significance. We further mapped chromatin interaction-based associations using H-MAGMA 72 across the 28 tissues and cell types of annotation of chromatin accessibility released with the software, applying default parameters. We used the individuals of AMR ancestry from the 1000 Genomes Project as the LD reference panel. To identify GWS gene-level associations, we applied a Bonferroni correction for the number of genes tested (P < 0.05/1,187,863 = 4.21x10 − 8 ). Brain Regulatory Chromatin Interactions To map SNPs with regulatory elements in the brain, we leveraged high-confidence regulatory chromatin interactions (HCRCI) identified from Hi-C datasets generated from adult (N = 3 temporal cortex) and fetal (N = 3 cortex) postmortem human brains 73 . Chromatin interactions were defined using 10 kb binds spaced 30 kb to 2 Mb apart, and interactions were considered high-confidence if they passed a Bonferroni-adjusted significance threshold (P < 0.005/42,985,244 = 1.16x10 − 10 ). We excluded ENCODE blacklist regions, centromeric regions, and low-quality bins. Regulatory chromatin interactions were defined by overlaps with annotated promoters or enhancers. SNPs were mapped to these interactions based on proximity (± 10 kb) to either anchor region of the HCRCIs. HCRC coordinates were lifted over from hg19 to hg38 using UCSC LiftOver 186 . We intersected SNP-HCRCI pairs with protein-coding genes expressed in the human brain based on GENCODE v45 187 gene annotations (geneMatrix dataset; curated by Patrick Sullivan, updated 03/2024). Genes located within 5 kb of an HCRCI anchor were considered linked to regulatory chromatin interactions. Analyses were conducted separately for adult and fetal cortex using R version 4.2. Transcriptome-wide association studies We performed transcriptome-wide association studies integrating both gene expression (eTWAS) and splicing isoforms (sTWAS) using S-PrediXcan 74 . We employed expression and splicing models trained with a multivariate adaptive shrinkage approach (Mashr) 188 from the Genotype-Tissue Expression (GTEx) project 189 – 191 . Prediction models and covariance matrices were obtained from the PredictDB repository ( http://predictdb.org/ ). Significance was determined via Bonferroni correction for the total number of gene–tissue pairs tested: eTWAS, P < 0.05/223,665 = 2.24x10 − 7 ; and sTWAS, P < 0.05/575,722 = 8.68x10 − 8 ). To improve power, we also applied S-MultiXcan 75 , which aggregates single-tissue TWAS signals across tissues. We used the same expression models for cross-tissue inference: eTWAS (S-MultiXcan): P < 0.05/17,642 = 2.83x10 − 6 ; and sTWAS (S-MultiXcan): P < 575,722 = 8.68x10 − 8 ). Additionally, we applied the FUSION framework 76 , a summary-based TWAS method that accounts for LD structure via a reference panel. We used FUSION prediction models trained on AMR individuals and calculated SNP weights using population-matched LD. Bonferroni correction was applied to all gene–tissue pairs: FUSION TWAS: P < 0.05/325,561 = 1.54x10 − 7 ). To further enhance cross-tissue signal detection, we employed the sCCA-ACAT method 77 , which integrates single-tissue TWAS statistics via sparse canonical correlation analysis combined with the aggregated Cauchy association test (P < 0.05/28,695 = 1.74x10 − 6 ). Proteome-wide association study (PWAS) To identify genetic associations reflected at the plasma protein level, we performed a proteome-wide association study using S-PrediXcan 74 and protein quantitative locus (pQTL) models trained to predict the plasma levels of 4,657 proteins in European American and African American individuals 78 , 79 . Genome-wide significance was defined using a Bonferroni correction for the number of tested proteins (P < 0.05/4,657 = 1.07x10 − 5 ). In addition, we analyzed models derived from 1,305 proteins measured in individuals of Hispanic descent 78 to capture ancestry-relevant protein associations. Multi-tissue single-cell trait risk To determine the risk of cell types across tissues in mediating genetic associations, we applied scPagwas 80 to integrate GWAS meta-analysis results with single-cell expression datasets from multiple human tissues. We included individuals without psychiatric disorders from publicly available single-cell datasets of brain tissues implicated in alcohol-related traits, including the amygdala 81 , striatum 82 , cerebellum 83 , hypothalamus 83 , hippocampus 83 , and cortex 83 . Peripheral tissues were also included: adipose tissue 84 , arteries 85 , and liver 86 . We included only individuals of EUR ancestry, given that current public single-cell repositories does not include individuals of Latin American descent. We first analyzed common variants with a MAF > 0.01. For loci showing cell type-specific association, we conducted secondary analysis including variants with MAF > 0.001, given that several genome-wide significant SNPs had low allele frequency. Genes were mapped to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways 192 in each cell type, and trait-risk scores were computed based on gene-pathway mapping. Variant coordinates were based on the hg38 human genome assembly and LD structure from the 1000 Genomes Project. We performed 100 iterations per cell type per tissue. Genes with consistent Pearson correlation coefficients (PCC > 0.1 or PCC < -0.1) across all tissues were further interrogated using Gene Ontology (GO) enrichment analysis. Tissue and pathway overlaps were visualized using the clusterProfiler R package 193 , leveraging the compareCluster function to assess shared biological processes. Network analysis We identified 261 unique genes across 123 gene prioritization methods applied to the alcohol consumption GWAS summary statistics. One method, scPagwas-Amygdala, was excluded from downstream analysis, as it uniquely prioritized 158 genes not supported by any other approach. This refinement yielded a working set of 103 genes identified by the remaining 122 methods. To integrate multiple sources of biological evidence, we constructed a 33-layer multiplex gene network comprising 56,218 unique Ensembl gene IDs 194 and 67,112,775 total edges. This network included five layers from HumanNet v3 195 (co-citation, co-expression, molecular pathway, gene interaction, and gene neighborhood networks), a dorsolateral prefrontal cortex (dlPFC)-specific transcription factor-target network 196 , and a protein-protein interaction network layer merged from HumanNet v3 and STRING 197 . The remaining 26 layers consisted of predictive expression networks (PENs). Among these, we included two bulk tissue-specific PENs from dlPFC and nucleus accumbens (NAc), generated from GTEx version 8 190 RNA-seq data using iRF-LOOP 198 . Additionally, we constructed 16 cell type-specific PENs (scPENs) from previously published single-nuclear RNA-seq (snRNA-seq) from the dlPFC 199 , representing diverse neuronal subtypes including ID2-expressing caudal ganglionic eminence (CGE)-derived inhibitory interneurons, LAMP5/NOS1-expressing CGE-derived inhibitory interneurons, VIP-expressing CGE-derived inhibitory interneurons, PV-expressing medial ganglionic eminence (MGE)-derived inhibitory interneurons, PV/SCUBE3-expressing MGE-derived inhibitory neurons, SST-expressing MGE-derived inhibitory interneurons, CUX2-expressing layer II/III principal excitatory neurons, RORB-expressing layer IV principal excitatory neurons, THEMIS-expressing layer V/VI principal excitatory neurons, TLE4-expressing layer V/VI principal excitatory neurons, astrocytes, microglia, oligodendrocytes, mature oligodendrocytes, oligodendrocyte precursor cells (OPCs), and vascular cells. We also included 9 scPENs from previously published snRNA-seq data from the NAc 83 CGE-derived interneurons, medium spiny neurons, eccentric medium spiny neurons, astrocytes, ependymal cells, microglia, oligodendrocytes, OPCs, and splatter cells. To refine the GWAS-prioritized gene set and reduce false positives, we applied Gene set Refinement using Interacting Networks (GRIN) 87 , which uses a Random Walk with Restart (RWR) algorithm to assess the interconnectivity of genes in the multiplex network. The connectivity of each gene is compared against a null distribution derived from random gene sets, and genes whose rankings deviate significantly from the null are retained. Of the 103 prioritized genes, 96 were found in the multiplex network, and GRIN analysis reduced this set to 31 high-confidence genes. We then used Multiplex Embedding of Networks for Team-based Omics Research (MENTOR) 88 to cluster these 31 GRIN-retained genes into mechanistic modules based on network topology MENTOR applies RWR from each gene to generate a rank-ordered embedding all other genes in the network, then calculates pairwise Spearman correlation coefficients (ρ) between these embeddings, which is then converted into a distance matrix by calculating min[1 - ρ, 1]. These distances are then clustered into a dendrogram using agglomerative hierarchical sub-clustering using complete linkage, ensuring each module reaches a maximum of 20 genes. This approach generated functionally coherent gene modules supported by multiple lines of evidence across network layers. A polar dendrogram was used to visualize these modules, and an accompanying heatmap displayed the contributions of the original data sources to each gene’s network connectivity. Genetic Correlations To assess shared genetic architecture, we downloaded and processed 64 GWAS summary statistics from the GWAS Catalog 89 that included Hispanic/Latin American populations ( Supplementary Table 29 ). Genetic correlations (rg) were estimated using cov-LDSC with the SIGMA reference panel 200 , which accounts for LD patterns in admixed populations. Correlations with P-values below 7.81x10 − 4 were considered statistically significant. Polygenic risk scores (PRS) To evaluate the transferability of PRS across Latin American groups, we calculated several PRS in the HCHS/SOL cohort. First, we generated a PRS based solely on Biobank cohorts (23andMe, MCPS, and MVP). Second, we computed a PRS using all cohorts from the fixed-effect meta-analysis, excluding HCHS/SOL. Third, we constructed a PRS using summary statistics from a previously published European ancestry GWAS 16 . PRS-CS 90 , a Bayesian regression framework that uses continuous shrinkage (CS) priors and incorporates ancestry-matched LD, was used to derive posterior effect size estimates for SNPs in HapMap3. For each PRS, we performed linear regression to test associations with drinks consumed per week (DrinksWk), adjusting for age, sex, and 10 PCs of ancestry calculated using PCAiR 201 . Both PRS and DrinksWk were standardized to have a mean of zero and a standard deviation of 1. Related individuals were removed using KING-robust as implemented in PCAiR. To evaluate predictive performance, we compared adjusted R 2 values from models with and without the PRS term 202 . We also assessed polygenic prediction across ancestries by applying PRS-CSx 91 , an extension of PRS-CS that jointly models GWAS summary statistics across populations by leveraging shared CS priors and ancestry-specific LD structures. This analysis included a fourth PRS derived from African ancestry summary statistics from the same prior study 16 . Population-specific PRSs were computed using posterior SNP effect sizes estimated by PRS-CSx. A combined multi-ancestry PRS was then created by fitting a linear model that integrated the three ancestry-specific PRSs (This Study, EUR, and AFR), each derived using a fixed global shrinkage parameter. Association models were repeated as described for PRS-CS. We analyzed the transferability of the PRS constructed using PRS-CS and PRSCSx in the HCHS/SOL cohort, stratifying the analysis based on the geographic origin (Central America, Cuba, Dominican, Mexican, Puerto Rican, and South American). Within each group, we fitted linear regression models relating DrinksWk to each PRS and the different PRSs adjusted by age, sex, and 10 principal components of ancestry, in each geographical subgroup. We next stratified analyses within HCHS/SOL to evaluate PRS performance by geographic origin, including individuals self-identifying as Central American, Cuban, Dominican, Mexican, Puerto Rican, or South American. Within each group, we fitted linear regression models relating DrinksWk to each PRS, adjusting for the same covariates. In a complementary approach, we used unsupervised clustering to explore genetic structure and its relationship to PRS transferability. Following a previously published machine learning approach in Latin American individuals from the MVP cohort 93 , we applied k-means clustering to HCHS/SOL genetic data, varying the number of clusters from 1 to 10 and incrementally adding PCs (2 to 32) as features. Clustering performance was evaluated using the Silhouette score 92 , which quantifies how well individuals fit within their assigned object to its assigned cluster compared to other clusters, with higher values indicating better-defined clusters. We then fitted stratified linear regression to test the association between DrinksWk and the different PRS’s adjusted by age, sex, and 10 principal components of ancestry, in each cluster identified by the machine learning algorithm. We analyzed the ancestry proportions of each cluster using Admixture 94 , using the 1000 Genomes and Human Genome Diversity Project 95 as reference panel. We additionally assessed the predictive performance of the PRS using PRS-CS 90 , in newly genotyped individuals from the MXGDAR-Freeze2 (n = 624), and two additional cohorts El Banco por Salud (n = 680), and EPISONO (n = 565), for two additional alcohol phenotypes. El Banco por Salud is a biobank that aims to advance the study of Type 2 Diabetes in Arizona residents of Mexican ancestry 48 . The São Paulo Epidemiologic Sleep Study (EPISONO) is a population-based study of sleep and risk factors associated with sleep disturbances 49 . For these cohorts, we removed related individuals using King-robust 203 implemented in PCAiR. We used only the HapMap3 SNPs overlapping across all cohorts for PRS calculations and standardized the PRS to have a mean of 0 and standard deviation of 1. In these cohorts, we performed correlation with the number of drinks consumed per occasion and frequency of consumption; we used the coding used in the AUDIT-C, and after standardizing the phenotypes. Then we ran linear models using the same covariates (age, sex, and 10 principal components of genetic ancestry). Declarations Acknowledgments The Spit for Science Working Group: Director: Karen Chartier. Co-Director: Ananda Amstadter. Past Founding Director: Danielle M. Dick (2011-2022). Registry management: Emily Lilley, Renolda Gelzinis, Anne Morris. Data cleaning and management: Katie Bountress, Amy E. Adkins, Nathaniel Thomas, Zoe Neale, Kimberly Pedersen, Thomas Bannard & Seung B. Cho. Data collection: Kimberly Pedersen, Amy E. Adkins, Peter Barr, Holly Byers, Erin C. Berenz, Erin Caraway, Seung B. Cho, James S. Clifford, Megan Cooke, Elizabeth Do, Alexis C. Edwards, Neeru Goyal, Laura M. Hack, Lisa J. Halberstadt, Sage Hawn, Sally Kuo, Emily Lasko, Jennifer Lent, Mackenzie Lind, Elizabeth Long, Alexandra Martelli, Jacquelyn L. Meyers, Kerry Mitchell, Ashlee Moore, Arden Moscati, Aashir Nasim, Zoe Neale, Jill Opalesky, Cassie Overstreet, A. Christian Pais, Tarah Raldiris, Jessica Salvatore, Jeanne Savage, Rebecca Smith, David Sosnowski, Jinni Su, Nathaniel Thomas, Chloe Walker, Marcie Walsh, Teresa Willoughby, Madison Woodroof & Jia Yan. Genotypic data processing and cleaning: Cuie Sun, Brandon Wormley, Brien Riley, Fazil Aliev, Roseann E. Peterson & Bradley T. Webb. We would like to thank the Spit for Science participants for making this study a success, as well as the many Universities faculty, students, and staff who contributed to the design and implementation of the project. This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan). Funding This work was funded by the National Institutes of Health (Psychiatric Genomic Consortium - Substance Use Disorder R01DA054869; R01HG012869 to EG. The work was also supported by the Kavli Institute for Neuroscience at Yale University, Kavli Postdoctoral Award for Academic Diversity to Jose Jaime Martinez-Magaña, by NIDA grants R21DA050160 and DP1DA058737 (JLMO, JJMM), and National Institute of Mental Health grant R01MH136157 (JJMM). VA MVP grant I01 BX004820 (HRK, ACJ) and IO1CX001849 (JG, HZ). FAPESP (#2020/13467-8 to MLA) and AFIP, CNPq to MLA and ST. MCPS was supported by the Mexican Health Ministry; the National Council of Science and Technology for Mexico; Wellcome [058299/Z/99]; Cancer Research UK; the British Heart Foundation [RE/13/1/30181]; and the UK Medical Research Council [MC_UU_00017/2, MR/Z504543/1]. Spit for Science has been supported by Virginia Commonwealth University, P20AA017828, R37AA011408, K02AA018755, P50AA022537, and K01AA024152 from the National Institute on Alcohol Abuse and Alcoholism, UL1RR031990 from the National Center for Research Resources and National Institutes of Health Roadmap for Medical Research, as well as support by the Center for the Study of Tobacco Products at VCU. REDCap support provided by CTSA award UM1TR004360 from the National Center for Advancing Translational Sciences. The content is solely the responsibility of the authors and does not necessarily represent the views of the respective funding agencies. RK was supported by NIAAA (K01 AA028292); VISN 4 Mental Illness Research, Education and Clinical Center; US Department of Veterans Affairs (I01 BX004820). MPP and ST was supported by CNPq 465550-2014-2, Fapesp 2021/05332-8 and 2021/12901-9. KT was supported by NIH P01 AG023394, P50 HL105185. MMO and PT was supported by Associação Fundo de Incentivo à Pesquisa (AFIP), FAPESP 2021/09089-0. AM was supported by Associação Fundo de Incentivo à Pesquisa (AFIP), CNPq, FAPESP 2020/13467-8. Conflict of interest HRK is a member of advisory boards for Altimmune and Clearmind Medicine; a consultant to Sobrera Pharmaceuticals, Altimmune, Lilly, and Ribocure; the recipient of research funding and medication supplies for an investigator-initiated study from Alkermes and a company-initiated study by Altimmune; and an inventor on U.S. provisional patent “Multi-ancestry Genome-wide Association Meta-analysis of Buprenorphine Treatment Response. LAR has received grant or research support from, served as a consultant to, and served on the speakers’ bureau of Abdi Ibrahim, Abbott, Aché, Adium, Apsen, Bial, Cellera, EMS, Hypera Pharma, Knight Therapeutics, Libbs, Medice, Novartis/Sandoz, Pfizer/Upjohn/Viatris, Shire/Takeda, and Torrent in the last three years. The ADHD and Juvenile Bipolar Disorder Outpatient Programs chaired by LAR have received unrestricted educational and research support from the following pharmaceutical companies in the last three years: Novartis/Sandoz and Shire/Takeda. LAR has received authorship royalties from Oxford Press and ArtMed. MCPS was supported by grants from Regeneron and AstraZeneca to the University of Oxford. Data availability The full summary-level association data from the meta-analyses are publicly available through XXXXXX. Data from the Mexico City Prospective Study are available to bona fide academic researchers. For more details, the study’s Data and Sample Sharing policy may be downloaded (in English or Spanish) from https://www.ctsu.ox.ac.uk/research/mcps. Available study data can be examined in detail through the study’s Data Showcase, available at https://datashare.ndph.ox.ac.uk/mexico/. MCPS ancestry-specific allele frequencies are available in a public browser (https://rgc-mcps.regeneron.com/). Data from the S4S study are available to qualified researchers via dbGaP (phs001754.v4.p2) or via [email protected] who provide the appropriate signed data use agreement. Code availability All software used in this study is publicly available TOPMed Imputation Server, https://imputation.biodatacatalyst.nhlbi.nih.gov/#!; GWASlab; PLINK; MTAG; METAL; GCTA; LDSC; cov-LDSC; MAGMA; H-MAGMA; Tractor; Shapeit5; RFMix2; PCAiR; S-PrediXcan; scPagwas, clusterProfiler, and S-MultiXcan, https://github.com/hakyimlab/MetaXcan; PRS-CS; PRS-Csx; Fusion; ACAT; SusieR; GRIN; MENTOR; Admixture; LiftOver; SNPNexus; PCAiR; KING; MENTOR; GRIN; Admixture References Rehm J et al (2003) The relationship of average volume of alcohol consumption and patterns of drinking to burden of disease: an overview. Addiction 98:1209–1228 Griswold MG et al (2018) Alcohol use and burden for 195 countries and territories, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 392:1015–1035 Gilmore W et al (2016) Alcohol: taking a population perspective. Nat Rev Gastroenterol Hepatol 13, 426 – 34 Cremonte M, Biscarra MA, Conde K, Cherpitel CJ (2018) Epidemiology of alcohol consumption and related problems in Latin American countries: Contributions of psychology. Int J Psychol 53:245–252 Diaz LA, Roblero JP, Bataller R, Arab JP (2020) Alcohol-Related Liver Disease in Latin America: Local Solutions for a Global Problem. Clin Liver Dis (Hoboken) 16:187–190 Trichia E et al (2024) Alcohol and mortality in Mexico: prospective study of 150 000 adults. Lancet Public Health 9:e907–e915 Diaz LA et al (2022) Liver Diseases in Latin America: Current Status, Unmet Needs, and Opportunities for Improvement. Curr Treat Options Gastroenterol 20:261–278 Huang DQ et al (2023) Global epidemiology of cirrhosis - aetiology, trends and predictions. Nat Rev Gastroenterol Hepatol 20:388–398 Arab JP, Bataller R, Roblero JP (2020) Are We Really Taking Care of Alcohol-Related Liver Disease in Latin America? Clin Liver Dis (Hoboken) 16:91–95 Roblero JP, Diaz LA, Arab JP, Roblero P, Bataller R (2021) Socioeconomic inequalities in the incidence of alcohol-related liver disease in the Latin American context. Lancet Reg Health Eur 10:100229 Sanchez-Roige S et al (2019) Genome-Wide Association Study Meta-Analysis of the Alcohol Use Disorders Identification Test (AUDIT) in Two Population-Based Cohorts. Am J Psychiatry 176:107–118 Kranzler HR et al (2019) Genome-wide association study of alcohol consumption and use disorder in 274,424 individuals from multiple populations. Nat Commun 10:1499 Lai D et al (2019) Genome-wide association studies of alcohol dependence, DSM-IV criterion count and individual criteria. Genes Brain Behav 18:e12579 Deak JD et al (2022) Genome-Wide Investigation of Maximum Habitual Alcohol Intake in US Veterans in Relation to Alcohol Consumption Traits and Alcohol Use Disorder. JAMA Netw Open 5:e2238880 Zhou H et al (2023) Multi-ancestry study of the genetics of problematic alcohol use in over 1 million individuals. Nat Med 29:3184–3192 Saunders GRB et al (2022) Genetic diversity fuels gene discovery for tobacco and alcohol use. Nature 612:720–724 Mallard TT et al (2022) Item-Level Genome-Wide Association Study of the Alcohol Use Disorders Identification Test in Three Population-Based Cohorts. Am J Psychiatry 179:58–70 Gelernter J et al (2019) Genome-wide Association Study of Maximum Habitual Alcohol Intake in > 140,000 U.S. European and African American Veterans Yields Novel Risk Loci. Biol Psychiatry 86:365–376 Wojcik GL et al (2019) Genetic analyses of diverse populations improves discovery for complex traits. Nature 570:514–518 Peterson RE et al (2019) Genome-wide Association Studies in Ancestrally Diverse Populations: Opportunities, Methods, Pitfalls, and Recommendations. Cell 179:589–603 Mills MC, Rahal C (2020) The GWAS Diversity Monitor tracks diversity by disease in real time. Nat Genet 52:242–243 Maldonado BL, Pique DG, Kaplan RC, Claw KG, Gignoux CR (2023) Genetic risk prediction in Hispanics/Latinos: milestones, challenges, and social-ethical considerations. J Community Genet 14:543–553 Kember RL et al (2023) Genetic Underpinnings of the Transition From Alcohol Consumption to Alcohol Use Disorder: Shared and Unique Genetic Architectures in a Cross-Ancestry Sample. Am J Psychiatry 180:584–593 Bruxel EM et al (2025) Psychiatric genetics in the diverse landscape of Latin American populations. Nat Genet 57:1074–1088 Wang S et al (2007) Genetic variation and population structure in native Americans. PLoS Genet 3:e185 Wang S et al (2008) Geographic patterns of genome admixture in Latin American Mestizos. PLoS Genet 4:e1000037 Reich D et al (2012) Reconstructing Native American population history. Nature 488:370–374 Ruiz-Linares A et al (2014) Admixture in Latin America: geographic structure, phenotypic diversity and self-perception of ancestry based on 7,342 individuals. PLoS Genet 10:e1004572 Homburger JR et al (2015) Genomic Insights into the Ancestry and Demographic History of South America. PLoS Genet 11:e1005602 Rodriguez-Rodriguez JE et al (2022) The genetic legacy of the Manila galleon trade in Mexico. Philos Trans R Soc Lond B Biol Sci 377:20200419 Norris ET et al (2018) Genetic ancestry, admixture and health determinants in Latin America. BMC Genomics 19:861 Aragones A, Hayes SL, Chen MH, Gonzalez J, Gany FM (2014) Characterization of the Hispanic or latino population in health research: a systematic review. J Immigr Minor Health 16, 429 – 39 Lorenzo-Blanco EI et al (2016) Alcohol use among recent immigrant Latino/a youth: acculturation, gender, and the Theory of Reasoned Action. Ethn Health 21:609–627 Roche KM et al (2021) Recent immigration actions and news and the adjustment of U.S. Latino/a adolescents. Cultur Divers Ethnic Minor Psychol 27:447–459 Sanchez M et al (2016) Drinking and Driving among Recent Latino Immigrants: The Impact of Neighborhoods and Social Support. Int J Environ Res Public Health 13 Vazquez V et al (2023) Alcohol use among adult recent Latino/a immigrants: the role of stress, forced migration, and adherence to traditional gender roles. Am J Drug Alcohol Abuse 49:216–227 Atkinson EG et al (2022) Cross-ancestry genomic research: time to close the gap. Neuropsychopharmacology 47:1737–1738 Martin AR et al (2019) Clinical use of current polygenic risk scores may exacerbate health disparities. Nat Genet 51:584–591 Martin AR et al (2022) Increasing diversity in genomics requires investment in equitable partnerships and capacity building. Nat Genet 54:740–745 Tapia-Conyer R et al (2006) Cohort profile: the Mexico City Prospective Study. Int J Epidemiol 35:243–249 Ziyatdinov A et al (2023) Genotyping, sequencing and analysis of 140,000 adults from Mexico City. Nature 622:784–793 Martinez-Magana JJ et al (2020) The Identification of Admixture Patterns Could Refine Pharmacogenetic Counseling: Analysis of a Population-Based Sample in Mexico. Front Pharmacol 11:324 Salum GA et al (2015) High risk cohort study for psychiatric disorders in childhood: rationale, design, methods and preliminary results. Int J Methods Psychiatr Res 24:58–73 Gaziano JM et al (2016) Million Veteran Program: A mega-biobank to study genetic influences on health and disease. J Clin Epidemiol 70, 214 – 23 All of Us Research Program (2019) The All of Us Research Program. N Engl J Med 381:668–676 Lavange LM et al (2010) Sample design and cohort selection in the Hispanic Community Health Study/Study of Latinos. Ann Epidemiol 20:642–649 Dick DM et al (2014) Spit for Science: launching a longitudinal study of genetic and environmental influences on substance use and emotional health at a large US university. Front Genet 5:47 Parra OD et al (2022) Biobanking in Latinos: current status, principles for conduct, and contribution of a new biobank, El Banco por Salud, designed to improve the health of Latino patients of Mexican ancestry with type 2 diabetes. BMJ Open Diabetes Res Care 10 Lucena L et al (2020) The association of insomnia and quality of life: Sao Paulo epidemiologic sleep study (EPISONO). Sleep Health 6:629–635 Geovanini GR et al (2019) Poor sleep quality and lipid profile in a rural cohort (The Baependi Heart Study). Sleep Med 57:30–35 Arellano-Campos O et al (2019) Development and validation of a predictive model for incident type 2 diabetes in middle-aged Mexican adults: the metabolic syndrome cohort. BMC Endocr Disord 19:41 Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA (1998) The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. Ambulatory Care Quality Improvement Project (ACQUIP). Alcohol Use Disorders Identification Test. Arch Intern Med 158:1789–1795 Willer CJ, Li Y, Abecasis GR (2010) METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26:2190–2191 Bulik-Sullivan BK et al (2015) LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet 47:291–295 de Vries PS et al (2019) Multiancestry Genome-Wide Association Study of Lipid Levels Incorporating Gene-Alcohol Interactions. Am J Epidemiol 188:1033–1054 Li D, Zhao H, Gelernter J (2012) Strong protective effect of the aldehyde dehydrogenase gene (ALDH2) 504lys (*2) allele against alcoholism and alcohol-induced medical diseases in Asians. Hum Genet 131:725–737 Croteau DL, Popuri V, Opresko PL, Bohr VA (2014) Human RecQ helicases in DNA repair, recombination, and replication. Annu Rev Biochem 83:519–552 Zhao B et al (2019) Genome-wide association analysis of 19,629 individuals identifies variants influencing regional brain volumes and refines their genetic co-architecture with cognitive and mental health traits. Nat Genet 51:1637–1644 Wang G, Sarkar A, Carbonetto P, Stephens M (2020) A simple new approach to variable selection in regression, with application to genetic fine mapping. J R Stat Soc Ser B Stat Methodol 82:1273–1300 Turley P et al (2018) Multi-trait analysis of genome-wide association summary statistics using MTAG. Nat Genet 50:229–237 Byrska-Bishop M et al (2022) High-coverage whole-genome sequencing of the expanded 1000 Genomes Project cohort including 602 trios. Cell 185:3426–3440e19 Gudmundsson S et al (2022) Variant interpretation using population databases: Lessons from gnomAD. Hum Mutat 43:1012–1030 Yasumizu Y et al (2020) Genome-Wide Natural Selection Signatures Are Linked to Genetic Risk of Modern Phenotypes in the Japanese Population. Mol Biol Evol 37:1306–1316 Han Y et al (2007) Evidence of positive selection on a class I ADH locus. Am J Hum Genet 80:441–456 Chen CH, Kraemer BR (2022) & Mochly-Rosen, D. ALDH2 variance in disease and populations. Dis Model Mech 15 Im PK et al (2022) Alcohol metabolism genes and risks of site-specific cancers in Chinese adults: An 11-year prospective study. Int J Cancer 150:1627–1639 Zhang J et al (2023) The role of aldehyde dehydrogenase 2 in cardiovascular disease. Nat Rev Cardiol 20:495–509 Oscanoa J et al (2020) SNPnexus: a web server for functional annotation of human genome sequence variation (2020 update). Nucleic Acids Res 48:W185–W192 Jacob AG, Smith (2017) C.W.J. Intron retention as a component of regulated gene expression programs. Hum Genet 136:1043–1057 Lukacisin M, Espinosa-Cantu A, Bollenbach T (2022) Intron-mediated induction of phenotypic heterogeneity. Nature 605:113–118 de Leeuw CA, Mooij JM, Heskes T, Posthuma D (2015) MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput Biol 11:e1004219 Sey NYA et al (2020) A computational tool (H-MAGMA) for improved prediction of brain-disorder risk genes by incorporating brain chromatin interaction profiles. Nat Neurosci 23:583–593 Lu L et al (2020) Robust Hi-C Maps of Enhancer-Promoter Interactions Reveal the Function of Non-coding Genome in Neural Development and Diseases. Mol Cell 79:521–534e15 Barbeira AN et al (2018) Exploring the phenotypic consequences of tissue specific gene expression variation inferred from GWAS summary statistics. Nat Commun 9:1825 Barbeira AN et al (2019) Integrating predicted transcriptome from multiple tissues improves association detection. PLoS Genet 15:e1007889 Gusev A et al (2016) Integrative approaches for large-scale transcriptome-wide association studies. Nat Genet 48, 245 – 52 Feng H et al (2021) Leveraging expression from multiple tissues using sparse canonical correlation analysis and aggregate tests improves the power of transcriptome-wide association studies. PLoS Genet 17:e1008973 Schubert R et al (2022) Protein prediction for trait mapping in diverse populations. PLoS ONE 17:e0264341 Zhang J et al (2022) Plasma proteome analyses in individuals of European and African ancestry identify cis-pQTLs and models for proteome-wide association studies. Nat Genet 54:593–602 Ma Y et al (2023) Polygenic regression uncovers trait-relevant cellular contexts through pathway activation transformation of single-cell RNA sequencing data. Cell Genom 3:100383 Yu B et al (2023) Molecular and cellular evolution of the amygdala across species analyzed by single-nucleus transcriptome profiling. Cell Discov 9:19 Phan BN et al (2024) Single nuclei transcriptomics in human and non-human primate striatum in opioid use disorder. Nat Commun 15:878 Siletti K et al (2023) Transcriptomic diversity of cell types across the adult human brain. Science 382:eadd7046 Tabula Sapiens C et al (2022) The Tabula Sapiens: A multiple-organ, single-cell transcriptomic atlas of humans. Science 376:eabl4896 Han X et al (2020) Construction of a human cell landscape at single-cell level. Nature 581:303–309 Guilliams M et al (2022) Spatial proteogenomics reveals distinct and evolutionarily conserved hepatic macrophage niches. Cell 185:379–396e38 Sullivan KA et al (2024) Analyses of GWAS signal using GRIN identify additional genes contributing to suicidal behavior. Commun Biol 7:1360 Sullivan KA et al (2024) MENTOR: Multiplex Embedding of Networks for Team-Based Omics Research. bioRxiv Sollis E et al (2023) The NHGRI-EBI GWAS Catalog: knowledgebase and deposition resource. Nucleic Acids Res 51:D977–D985 Ge T, Chen CY, Ni Y, Feng YA, Smoller JW (2019) Polygenic prediction via Bayesian regression and continuous shrinkage priors. Nat Commun 10:1776 Ruan Y et al (2022) Improving polygenic prediction in ancestrally diverse populations. Nat Genet 54:573–580 Lovmar L, Ahlford A, Jonsson M, Syvanen AC (2005) Silhouette scores for assessment of SNP genotype clusters. BMC Genomics 6:35 Clarke SL et al (2022) Race and Ethnicity Stratification for Polygenic Risk Score Analyses May Mask Disparities in Hispanics. Circulation 146:265–267 Alexander DH, Novembre J, Lange K (2009) Fast model-based estimation of ancestry in unrelated individuals. Genome Res 19:1655–1664 Koenig Z et al (2024) A harmonized public resource of deeply sequenced diverse human genomes. bioRxiv Barry CS et al (2023) How to estimate heritability: a guide for genetic epidemiologists. Int J Epidemiol 52:624–632 Huang J, Kleman N, Basu S, Shriver MD, Zaidi AA (2025) Interpreting SNP heritability in admixed populations. bioRxiv Bierut LJ et al (2012) ADH1B is associated with alcohol dependence and alcohol consumption in populations of European and African ancestry. Mol Psychiatry 17:445–450 Clarke TK et al (2017) Genome-wide association study of alcohol consumption and genetic overlap with other health-related traits in UK Biobank (N = 112 117). Mol Psychiatry 22:1376–1384 Sanchez-Roige S et al (2019) Genome-wide association study of alcohol use disorder identification test (AUDIT) scores in 20 328 research participants of European ancestry. Addict Biol 24:121–131 Walters RK et al (2018) Transancestral GWAS of alcohol dependence reveals common genetic underpinnings with psychiatric disorders. Nat Neurosci 21:1656–1669 Liu M et al (2019) Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use. Nat Genet 51:237–244 Evangelou E et al (2019) New alcohol-related genes suggest shared genetic mechanisms with neuropsychiatric disorders. Nat Hum Behav 3:950–961 Jennings MV et al (2024) A phenome-wide association and Mendelian randomisation study of alcohol use variants in a diverse cohort comprising over 3 million individuals. EBioMedicine 103:105086 Ehlers CL, Liang T, Gizer IR (2012) ADH and ALDH polymorphisms and alcohol dependence in Mexican and Native Americans. Am J Drug Alcohol Abuse 38:389–394 Sanchez-Roige S, Palmer AA, Clarke TK (2020) Recent Efforts to Dissect the Genetic Basis of Alcohol Use and Abuse. Biol Psychiatry 87:609–618 Savage JE et al (2024) Genetic Heterogeneity Across Dimensions of Alcohol Use Behaviors. Am J Psychiatry 181:1006–1017 Farrelly D et al (1999) Mice mutant for glucokinase regulatory protein exhibit decreased liver glucokinase: a sequestration mechanism in metabolic regulation. Proc Natl Acad Sci U S A 96:14511–14516 Lin WY (2022) Lifestyle Factors and Genetic Variants on 2 Biological Age Measures: Evidence From 94 443 Taiwan Biobank Participants. J Gerontol Biol Sci Med Sci 77:1189–1198 Vaxillaire M et al (2008) The common P446L polymorphism in GCKR inversely modulates fasting glucose and triglyceride levels and reduces type 2 diabetes risk in the DESIR prospective general French population. Diabetes 57:2253–2257 Legaki E et al (2022) Association of Alcohol Use Disorder Risk With ADH1B, DRD2, FAAH, SLC39A8, GCKR, and PDYN Genetic Polymorphisms. Vivo 36:2092–2104 Khan Y et al (2024) Combining Transdiagnostic and Disorder-Level GWAS Enhances Precision of Psychiatric Genetic Risk Profiles in a Multi-Ancestry Sample. medRxiv Pasman JA et al (2018) GWAS of lifetime cannabis use reveals new risk loci, genetic overlap with psychiatric traits, and a causal influence of schizophrenia. Nat Neurosci 21:1161–1170 Zhou H et al (2020) Genome-wide meta-analysis of problematic alcohol use in 435,563 individuals yields insights into biology and relationships with other traits. Nat Neurosci 23:809–818 Yan X et al (2018) Cadm2 regulates body weight and energy homeostasis in mice. Mol Metab 8:180–188 Nagel M, Watanabe K, Stringer S, Posthuma D, van der Sluis (2018) Item-level analyses reveal genetic heterogeneity in neuroticism. Nat Commun 9:905 Hill WD et al (2020) Genetic contributions to two special factors of neuroticism are associated with affluence, higher intelligence, better health, and longer life. Mol Psychiatry 25:3034–3052 Morris J et al (2019) Genetic variation in CADM2 as a link between psychological traits and obesity. Sci Rep 9:7339 Arends RM et al (2021) Associations between the CADM2 gene, substance use, risky sexual behavior, and self-control: A phenome-wide association study. Addict Biol 26:e13015 Pasman JA et al (2022) The CADM2 Gene and Behavior: A Phenome-Wide Scan in UK-Biobank. Behav Genet 52:306–314 Sanchez-Roige S et al (2023) CADM2 is implicated in impulsive personality and numerous other traits by genome- and phenome-wide association studies in humans and mice. Transl Psychiatry 13:167 Yin B, Wang X, Huang T, Jia J (2022) Shared Genetics and Causality Between Decaffeinated Coffee Consumption and Neuropsychiatric Diseases: A Large-Scale Genome-Wide Cross-Trait Analysis and Mendelian Randomization Analysis. Front Psychiatry 13:910432 Koller D et al (2024) Genetic contribution to the comorbidity between attention-deficit/hyperactivity disorder and substance use disorders. Psychiatry Res 333:115758 Schumann G et al (2016) KLB is associated with alcohol drinking, and its gene product beta-Klotho is necessary for FGF21 regulation of alcohol preference. Proc Natl Acad Sci U S A 113:14372–14377 Zhou H et al (2022) Genome-wide meta-analysis of alcohol use disorder in East Asians. Neuropsychopharmacology 47:1791–1797 Quillen EE et al (2014) ALDH2 is associated to alcohol dependence and is the major genetic determinant of daily maximum drinks in a GWAS study of an isolated rural Chinese sample. Am J Med Genet B Neuropsychiatr Genet 165B, 103 – 10 Zhang X, Sun A, Ge J (2021) Origin and Spread of the ALDH2 Glu504Lys Allele. Phenomics 1:222–228 Li H et al (2009) Refined geographic distribution of the oriental ALDH2*504Lys (nee 487Lys) variant. Ann Hum Genet 73:335–345 Ehlers CL (2007) Variations in ADH and ALDH in Southwest California Indians. Alcohol Res Health 30, 14 – 7 Crawford JE et al (2017) Natural Selection on Genes Related to Cardiovascular Health in High-Altitude Adapted Andeans. Am J Hum Genet 101:752–767 Castaldelli-Maia JM et al (2021) Exploring the Role of Alcohol Metabolizing Genotypes in a 12-Week Clinical Trial of Naltrexone for Alcohol Use Disorder. Biomolecules 11 Gross ER et al (2015) A personalized medicine approach for Asian Americans with the aldehyde dehydrogenase 2*2 variant. Annu Rev Pharmacol Toxicol 55:107–127 Huggett SB, Ikeda AS, Yuan Q, Benca-Bachman CE, Palmer RHC (2023) Genome- and transcriptome-wide splicing associations with alcohol use disorder. Sci Rep 13:3950 Li R et al (2023) RNA alternative splicing impacts the risk for alcohol use disorder. Mol Psychiatry 28:2922–2933 Kang J et al (2024) Whole exome sequencing analysis identifies genes for alcohol consumption. Nat Commun 15:5777 Rivero-Rios P et al (2023) Recruitment of the SNX17-Retriever recycling pathway regulates synaptic function and plasticity. J Cell Biol 222 Garcia-Perez R et al (2023) The landscape of expression and alternative splicing variation across human traits. Cell Genom 3:100244 Dilly GA, Kittleman CW, Kerr TM, Messing RO, Mayfield RD (2022) Cell-type specific changes in PKC-delta neurons of the central amygdala during alcohol withdrawal. Transl Psychiatry 12:289 Roberto M, Kirson D, Khom S (2021) The Role of the Central Amygdala in Alcohol Dependence. Cold Spring Harb Perspect Med 11 Aroor AR, Shukla SD (2004) MAP kinase signaling in diverse effects of ethanol. Life Sci 74:2339–2364 Agoglia AE et al (2015) Alcohol alters the activation of ERK1/2, a functional regulator of binge alcohol drinking in adult C57BL/6J mice. Alcohol Clin Exp Res 39:463–475 Zhao X et al (2025) Single-cell landscape of the intrahepatic ecosystem in alcohol-related liver disease. Clin Transl Med 15:e70198 Arikkath J et al (2009) Delta-catenin regulates spine and synapse morphogenesis and function in hippocampal neurons during development. J Neurosci 29:5435–5442 Mizutani T, Ishizaka A, Furuichi Y (2015) The Werner Protein Acts as a Coactivator of Nuclear Factor kappaB (NF-kappaB) on HIV-1 and Interleukin-8 (IL-8) Promoters. J Biol Chem 290:18391–18399 Yu T et al (2014) The regulatory role of activating transcription factor 2 in inflammation. Mediators Inflamm 950472 (2014) Tsujimoto K, Takamatsu H, Kumanogoh A (2023) The Ragulator complex: delving its multifunctional impact on metabolism and beyond. Inflamm Regen 43:28 Zhou LY et al (2023) Polygenic risk scores and kidney traits in the Hispanic/Latino population: The Hispanic Community Health Study/Study of Latinos. HGG Adv 4:100177 McArdle CE et al (2021) Findings from the Hispanic Community Health Study/Study of Latinos on the Importance of Sociocultural Environmental Interactors: Polygenic Risk Score-by-Immigration and Dietary Interactions. Front Genet 12:720750 Breeze CE, Beck S, Berndt SI, Franceschini N (2022) The missing diversity in human epigenomic studies. Nat Genet 54:737–739 Husquin LT et al (2018) Exploring the genetic basis of human population differences in DNA methylation and their causal impact on immune gene regulation. Genome Biol 19:222 Fraser HB, Lam LL, Neumann SM, Kobor MS (2012) Population-specificity of human DNA methylation. Genome Biol 13:R8 Abdellaoui A, Dolan CV, Verweij KJH, Nivard MG (2022) Gene-environment correlations across geographic regions affect genome-wide association studies. Nat Genet 54:1345–1354 Dick DM (2011) Gene-environment interaction in psychological traits and disorders. Annu Rev Clin Psychol 7:383–409 Gonzalez S et al (2021) Consent for Use of Genetic Data among US Hispanics/Latinos: Results from the Hispanic Community Health Study/ Study of Latinos. Ethn Dis 31:547–558 Lin DY et al (2014) Genetic association analysis under complex survey sampling: the Hispanic Community Health Study/Study of Latinos. Am J Hum Genet 95, 675 – 88 Consortium TGP et al (2010) A map of human genome variation from population-scale sequencing. Nature 467:1061–1073 Taliun D et al (2021) Sequencing of 53,831 diverse genomes from the NHLBI TOPMed Program. Nature 590:290–299 Yang J, Lee SH, Goddard ME, Visscher PM (2011) GCTA: a tool for genome-wide complex trait analysis. Am J Hum Genet 88:76–82 Tucker KL et al (2010) The Boston Puerto Rican Health Study, a longitudinal cohort study on health disparities in Puerto Rican adults: challenges and opportunities. BMC Public Health 10:107 Loh PR et al (2016) Reference-based phasing using the Haplotype Reference Consortium panel. Nat Genet 48:1443–1448 Fuchsberger C, Abecasis GR, Hinds DA (2015) minimac2: faster genotype imputation. Bioinformatics 31:782–784 Marchini J, Howie B, Myers S, McVean G, Donnelly P (2007) A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet 39:906–913 Fang H et al (2019) Harmonizing Genetic Ancestry and Self-identified Race/Ethnicity in Genome-wide Association Studies. Am J Hum Genet 105:763–772 Chang CC et al (2015) Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience 4:7 All of Us Research Program Genomics (2024) Genomic data in the All of Us Research Program. Nature 627:340–346 Conomos MP, Reiner AP, Weir BS, Thornton TA (2016) Model-free Estimation of Recent Genetic Relatedness. Am J Hum Genet 98:127–148 Krissaane I et al (2020) Scalability and cost-effectiveness analysis of whole genome-wide association studies on Google Cloud Platform and Amazon Web Services. J Am Med Inf Assoc 27:1425–1430 Sorlie PD et al (2010) Design and implementation of the Hispanic Community Health Study/Study of Latinos. Ann Epidemiol 20:629–641 Gallo LC et al (2014) The Hispanic Community Health Study/Study of Latinos Sociocultural Ancillary Study: sample, design, and procedures. Ethn Dis 24:77–83 Castaneda SF et al (2019) Alcohol use, acculturation and socioeconomic status among Hispanic/Latino men and women: The Hispanic Community Health Study/Study of Latinos. PLoS ONE 14:e0214906 Conomos MP et al (2016) Genetic Diversity and Association Studies in US Hispanic/Latino Populations: Applications in the Hispanic Community Health Study/Study of Latinos. Am J Hum Genet 98:165–184 Howie B, Fuchsberger C, Stephens M, Marchini J, Abecasis GR (2012) Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nat Genet 44:955–959 Harris PA et al (2009) Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inf 42:377–381 Harris PA et al (2019) The REDCap consortium: Building an international community of software platform partners. J Biomed Inf 95:103208 Peterson RE et al (2017) The utility of empirically assigning ancestry groups in cross-population genetic studies of addiction. Am J Addict 26:494–501 Delaneau O, Zagury JF, Marchini J (2013) Improved whole-chromosome phasing for disease and population genetic studies. Nat Methods 10:5–6 Han B, Eskin E (2011) Random-effects model aimed at discovering associations in meta-analysis of genome-wide association studies. Am J Hum Genet 88:586–598 International HapMap C et al (2010) Integrating common and rare genetic variation in diverse human populations. Nature 467, 52 – 8 Delaneau O, Howie B, Cox AJ, Zagury JF, Marchini J (2013) Haplotype estimation using sequencing reads. Am J Hum Genet 93:687–696 He Y, Masaru K, Yuka S, Yoichiro K (2023) GWASLab: a Python package for processing and visualizing GWAS summary statistics. Jxiv Yang J et al (2012) Conditional and joint multiple-SNP analysis of GWAS summary statistics identifies additional variants influencing complex traits. Nat Genet 44:369–375 S1-3 Hofmeister RJ, Ribeiro DM, Rubinacci S, Delaneau O (2023) Accurate rare variant phasing of whole-genome and whole-exome sequencing data in the UK Biobank. Nat Genet 55:1243–1249 Maples BK, Gravel S, Kenny EE, Bustamante CD (2013) RFMix: a discriminative modeling approach for rapid and robust local-ancestry inference. Am J Hum Genet 93:278–288 Atkinson EG et al (2021) Tractor uses local ancestry to enable the inclusion of admixed individuals in GWAS and to boost power. Nat Genet 53:195–204 Luo Y et al (2021) Estimating heritability and its enrichment in tissue-specific gene sets in admixed populations. Hum Mol Genet 30:1521–1534 Genovese G et al (2024) BCFtools/liftover: an accurate and comprehensive tool to convert genetic variants across genome assemblies. Bioinformatics 40 Frankish A et al (2023) GENCODE: reference annotation for the human and mouse genomes in 2023. Nucleic Acids Res 51:D942–D949 Urbut SM, Wang G, Carbonetto P, Stephens M (2019) Flexible statistical methods for estimating and testing effects in genomic studies with multiple conditions. Nat Genet 51:187–195 Garrido-Martin D, Borsari B, Calvo M, Reverter F, Guigo R (2021) Identification and analysis of splicing quantitative trait loci across multiple tissues in the human genome. Nat Commun 12:727 The GC et al (2015) The Genotype-Tissue Expression (GTEx) pilot analysis: Multitissue gene regulation in humans. Science 348:648–660 Lonsdale J et al (2013) The Genotype-Tissue Expression (GTEx) project. Nat Genet 45:580–585 Kanehisa M, Goto S (2000) KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28:27–30 Yu G, Wang LG, Han Y, He QY (2012) clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16:284–287 Birney E et al (2004) An overview of Ensembl. Genome Res 14:925–928 Kim CY et al (2022) HumanNet v3: an improved database of human gene networks for disease research. Nucleic Acids Res 50:D632–D639 Pearl JR et al (2019) Genome-Scale Transcriptional Regulatory Network Models of Psychiatric and Neurodegenerative Disorders. Cell Syst 8:122–135e7 Szklarczyk D et al (2019) STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res 47:D607–D613 Cliff A et al (2019) A High-Performance Computing Implementation of Iterative Random Forest for the Creation of Predictive Expression Networks. Genes (Basel) 10 Herring CA et al (2022) Human prefrontal cortex gene regulatory dynamics from gestation to adulthood at single-cell resolution. Cell 185:4428–4447e28 Williams AL et al (2014) Sequence variants in SLC16A11 are a common risk factor for type 2 diabetes in Mexico. Nature 506:97–101 Conomos MP, Miller MB, Thornton TA (2015) Robust inference of population structure for ancestry prediction and correction of stratification in the presence of relatedness. Genet Epidemiol 39:276–293 Piepho HP (2023) An adjusted coefficient of determination (R(2)) for generalized linear mixed models in one go. Biom J 65:e2200290 Manichaikul A et al (2010) Robust relationship inference in genome-wide association studies. Bioinformatics 26:2867–2873 Additional Declarations Yes there is potential Competing Interest. HRK is a member of advisory boards for Altimmune and Clearmind Medicine; a consultant to Sobrera Pharmaceuticals, Altimmune, Lilly, and Ribocure; the recipient of research funding and medication supplies for an investigator-initiated study from Alkermes and a company-initiated study by Altimmune; and an inventor on U.S. provisional patent “Multi-ancestry Genome-wide Association Meta-analysis of Buprenorphine Treatment Response. LAR has received grant or research support from, served as a consultant to, and served on the speakers’ bureau of Abdi Ibrahim, Abbott, Aché, Adium, Apsen, Bial, Cellera, EMS, Hypera Pharma, Knight Therapeutics, Libbs, Medice, Novartis/Sandoz, Pfizer/Upjohn/Viatris, Shire/Takeda, and Torrent in the last three years. The ADHD and Juvenile Bipolar Disorder Outpatient Programs chaired by LAR have received unrestricted educational and research support from the following pharmaceutical companies in the last three years: Novartis/Sandoz and Shire/Takeda. LAR has received authorship royalties from Oxford Press and ArtMed. MCPS was supported by grants from Regeneron and AstraZeneca to the University of Oxford. Supplementary Files 09SupplementaryTables12032025.xlsx Supplementary Tables 08SupplementaryFigures07172025.docx Supplementary Figures Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• Amerindian ancestry influences genetic susceptibility to chronic obstructive pulmonary disease#948922
• Nutritional Risk Factors for FASD in Ukraine#1113000
• Development of the Alcohol-Linked Typical Experienced Responses Scale#1054924
• Development of the Alcohol-Linked Typical Experienced Responses Scale#1054812
• Archetypal Analysis for population genetics#1134227