Dietary intake, a major contributor to the global obesity epidemic 1-5 , is a complex phenotype partially affected by innate physiological processes. [6][7][8][9][10][11] However, previous genome-wide association studies (GWAS) have only implicated a few loci in variability of dietary composition. 12-14 Here, we present a multi-trait genome-wide association meta-analysis of inter-individual variation in dietary intake in 283,119 European-ancestry participants from UK Biobank and CHARGE consortium, and identify 96 genome-wide significant loci. Dietary intake signals map to different brain tissues and are enriched for genes expressed in b1tanycytes and serotonergic and GABAergic neurons. We also find enrichment of biological pathways related to neurogenesis. Integration of cell-line and brainspecific epigenomic annotations identify 15 additional loci. Clustering of genomewide significant variants yields three main genetic clusters with distinct associations with obesity and type 2 diabetes (T2D). Overall, these results enhance biological understanding of dietary composition, highlight neural mechanisms, and support functional follow-up experiments.As dietary components are strongly correlated, we conducted a multi-trait genomewide association meta-analysis of overall variation in dietary intake among 283,119European-ancestry participants from the UK Biobank 15 and the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium 14 (Methods; Supplementary Table 1). First, we conducted single-trait GWAS for the proportion of total energy intake from carbohydrate, fat, and protein in UK Biobank (n=192,005).Next, single-trait GWAS from the UK Biobank and CHARGE Consortium (n=91,114) were meta-analyzed and combined into a multi-trait genome-wide association metaanalysis (Methods). An analysis overview is presented in Supplementary Fig. 1.We evaluated dietary intake using 24-hour web-based diet recall in the UK Biobank 16,17 and validated food frequency questionnaires, diet history and diet records in the CHARGE Consortium. 14 We observed strong genome-wide genetic correlations for nutrient estimates between the UK Biobank and CHARGE datasets (r g >0.6 for all; P <0.001; Supplementary Table 2). The quantile-quantile plots of single-trait and multi-trait meta-analyses showed moderate inflation (l GC ranging from 1.12 to 1.17) with a linkage disequilibrium (LD) score intercept 18 of ~1 (standard error (s.e.) = 0.01), indicating that most inflation could be explained by polygenic signal ( Supplementary Fig. 2, Supplementary Table 3). In single-trait meta-analyses, genome-wide SNP-based heritability 19 was estimated at 3.9% (s.e.=0.01), 2.8% (s.e.=0.01), and 3.0% (s.e.=0.01) for carbohydrate, fat, and protein, respectively ( Supplementary Table 3), in line with previous GWAS findings 12,14 and other behavioral phenotypes such as tobacco or alcohol use. 20