Inferring human neutral genetic variation from craniodental phenotypes

Abstract

There is a growing consensus that global patterns of modern human cranial and dental variation are shaped largely by neutral evolutionary processes, suggesting that craniodental features can be used as reliable proxies for inferring population structure and history in bioarchaeological, forensic, and paleoanthropological contexts. However, there is disagreement on whether certain types of data preserve a neutral signature to a greater degree than others. Here, we address this unresolved question and systematically test the relative neutrality of four standard metric and nonmetric craniodental data types employing an extensive computational genotype-phenotype comparison across modern populations from around the world. Our computation draws on the largest existing data sets currently available, while accounting for geographically structured environmental variation, population sampling uncertainty, disparate numbers of phenotypic variables, and stochastic variation inherent to a neutral model of evolution. Our results reveal that the four data types differentially capture neutral genomic variation, with highest signals preserved in dental nonmetric and cranial metric data, followed by cranial nonmetric and dental metric data. Importantly, we demonstrate that combining all four data types together maximizes the neutral genetic signal compared with using them separately, even with a limited number of phenotypic variables. We hypothesize that this reflects a lower level of genetic integration through pleiotropy between, compared to within, the four data types, effectively forming four different modules associated with relatively independent sets of loci. Therefore, we recommend that future craniodental investigations adopt holistic combined data approaches, allowing for more robust inferences about underlying neutral genetic variation.

Keywords: cranium; dentition; genetic variation; neutral evolution.

Fig. 1.

Geographic location and genomic and phenotypic relationships of worldwide modern human populations. (A) World map showing the locations of 26 populations sampled for matched genomic data (SNPs) and phenotypic data (cranial and dental metrics and nonmetric traits). Points are approximate geographic coordinates of the genomic samples. 3D NMDS plots of between-population distances, calculated separately from six different data types: (B) SNPs; (C) cranial metrics; (D) dental metrics; (E) cranial nonmetric traits; (F) dental nonmetric traits; and (G) combined craniodental data.

Fig. 2.

Violin plots showing neutrality estimates for five craniodental data types, calculated as partial Pearson correlation (r) between phenotypic (D²) and neutral genetic (F_ST) distances across 26 modern human population samples, controlling for climate distances. Box plots are superimposed to show median values (black solid line) and interquartile ranges (boxes). (A) Distribution of 1,000 iteratively generated r values, each iteration leaving out a randomly selected population in the phenotypic and genomic data sets and a randomly selected individual in each remaining population. (B) Distribution of 1,000 iteratively generated r values, each iteration randomly undersampling the number of phenotypic variables, combined with population sampling bias correction. (C) Distribution of 1,000 iteratively generated r values, each iteration randomly undersampling the number of loci, combined with population and phenotype sampling bias correction.