Genome-wide association study of diet composition identifies 21 loci and uncovers associations with health and behavior

The effect of diet coposition on weight, health, and longevity has been an important topic in epidemiology for decades. Yet, measuring diet composition accurately and identifying its causal effects remains a notoriously... [ view full abstract ]

The effect of diet coposition on weight, health, and longevity has been an important topic in epidemiology for decades. Yet, measuring diet composition accurately and identifying its causal effects remains a notoriously difficult task, and despite the lack of strong empirical evidence, different dietary guidelines have been published – with their focus recently shifting from recommending low-fat to low-carbohydrate (and in particular, low-sugar) diets. One aspect that is still largely missing from the literature are insights into the genetic architecture of diet composition and how it relates to lifestyle behaviors such as physical activity and health outcomes such as BMI, diabetes, and cardiovascular disease. Unobserved genetic effects could confound the observed phenotypic relationships between dietary intake and health, and insights into the genetic architecture of diet composition could help to model the effects of diet intake on health more accurately. Therefore, we conducted the largest genome-wide association study (GWAS) on diet composition to date.

In particular, our measure of diet composition was represented by four phenotypes, which broadly represented energy intake from fat, protein, carbohydrate, and sugar. These phenotypes were corrected for total caloric intake in a non-linear fashion, and were created on the basis of self-report questionnaire data on hundreds of food items in 18 different genotyped samples. With a combined maximum total sample size of N ~ 265k for fat, protein and carbohydrate, and N ~ 234k for sugar, we discover 21 unique independent loci across phenotypes, revealing surprising genetic associations with health outcomes and indicators of socioeconomic status and lifestyle. Finally, we find that the genetic signal for all diet composition phenotypes is particularly driven by SNPs expressed in the brain.