Determining how genetic variants lead to a particular phenotype is a real challenge. Most common traits and diseases, including type 2 diabetes and major depression, are influenced by thousands of genetic variants in the genome, each with a small effect on the trait under study. Now at eLife, Anil Raj (Calico Life Sciences), Gary Churchill (Jackson Laboratory) and their colleagues, and Kevin Wright report that the effect of a genetic variant on body weight in mice may depend on diet and age.

Body mass is a quantitative trait that can be easily measured in hundreds of mice, making it easier to detect subtle changes. It is also suitable for studying age effects, as mice can be weighed every week of their lives and the procedure has no effect on the animal. To investigate these effects, Wright et al. weighed 960 female mice once a week between 60 and 660 days of age. The mice were given unlimited food (fed at will) until they were 180 days old, and then randomly allocated to one of five dietary groups: fed at will, 20% caloric restriction, 40% caloric restriction, starve one day per week, or starve two days per week.

The researchers then tried to identify individual genetic loci that affect body weight and investigated the relationship between genetic effects and diet. They identified five loci with effects that were independent of diet alone and ten with effects that were dependent on diet alone. This discovery demonstrates that failure to model diet-dependent effects will result in the absence of genetic loci important for determining body weight. In addition, the team also identified nine loci with both independent and diet-dependent effects, raising the question of whether the same haplotype groups cause two types of effects. They found that different haplotype groups caused diet-independent effects in all cases. This supports the view that different genetic effects determine diet-independent differences in body mass in the general population and diet-dependent differences within dietary groups in terms of the specific DNA code.

The findings of Wright et al emphasise the need for contextualisation in genetic research. This need has been recognised for some time, but is difficult to meet because of uncertainty about what the appropriate context might be for a particular characteristic. Moreover, the statistical models required to account for context are becoming increasingly complex, requiring ever larger sample sizes.