Today, we know about 30% of the genes that cause variation in height in modern-day Europeans. By assuming that these genes had the same effect on height in ancient populations, we are able to use ancient DNA to predict the heights of humans who lived thousands of years ago. In almost all cases, however, the true stature of a skeleton is different from its predicted genetic height due to a number of different factors, such as nutrition and environmental changes, as found in a study conducted by Stephanie Marciniak. In a paper she published in 2021, our research mentor Samantha Cox was able to demonstrate that ancient genetic data is able to broadly predict height variation in ancient populations. Over the past few months, one of our jobs was to increase the sample size of this initial study by matching ancient genomes to their corresponding heights for each genetic sample we had gathered.

We also ran the same regression as in Cox’s 2021 paper on our new genetic and osteological data to observe changes in the predictive power of PRS, which is a score relating to how tall a human should be, generated from their genome. The R2 of the regression did decrease slightly from 0.063 to 0.044, but these values were both represented by significant p-values, so it appears that PRS does still accurately contribute to stature predictions.

We then ran the same regression that appeared in Marciniak’s 2022 paper and plotted the residuals by time period. We were able to replicate their results and show that the average residual in the Neolithic period was, in fact, negative. We then added sex as a covariate in the regression and noted that both males and females in the Neolithic period had nearly identical average residuals, which could suggest that this trend is being caused by a genetic shift rather than environmental stress, since environmental changes tend to affect the sexes differently.

Our final investigation was into changes in the effect sizes of the 1000 most significant genetic variants affecting height over time. We calculated effect sizes based off of the ancient data by using the slopes of the regressions on the relationship between femur length and the presence of each variant. We then compared ancient and modern effect sizes for each variant and removed any outliers that had fewer than 5 occurrences of either allele. We observed a strong positive correlation in the top 1000, 100, and 10 variants, so it appears that effect sizes have not changed much between our ancient and modern datasets, so predicting ancient statures using PRS based upon modern data should not be an issue.