Diet as a driver of natural selection in human evolution: a paleo-empirical perspective

Multiple genetic variants associated with diet-related traits show strong signatures of natural selection1-17. To test whether these signals were indeed diet-driven, we conducted an empirical investigation. We compiled an isotopic dataset comprising 6,064 ancient human samples and 5,635 food resource samples from Britain. We developed a Bayesian mixing model to estimate individual dietary proportions based on isotopic data and subsequently constructed a temporal dietary model. A dairy-use time series was also constructed18. Using 1,038 ancient DNA samples, we reconstructed derived allele frequency trajectories for 14 strongly selected single nucleotide polymorphisms (SNPs) via bootstrap resampling19. A generalized additive model (GAM) was then applied to estimate mean and time-varying selection coefficients, while accounting for evolutionary forces beyond selection. Finally, we applied the convergent cross mapping (CCM)20 algorithm for causal discovery between the time-varying selection coefficients and their corresponding dietary variables. Our findings indicate that C3 plant consumption drove selection on rs12401678 and rs653178, and dairy consumption on rs4988235. Selection signals at rs174570 and rs174594 are likely linked to marine fish and terrestrial meat intake, whereas the remaining SNPs show more complex selection dynamics in which any diet-related signal cannot be clearly identified. Our results underscore the complexity of natural selection at the genetic level and highlight the need for more careful evaluation when identifying its potential drivers.