PREDICTION OF EYE, HAIR AND SKIN COLOUR IN LATIN AMERICANS

PALMAL, SAGNIK; ADHIKARI, KAUSTUBH; PAROLIN, MARIA-LAURA; RUIZ-LINARES, ANDRÉS

FORENSIC SCIENCE INTERNATIONAL-GENETICS

ELSEVIER IRELAND LTD

Año: 2021

1872-4973

Here we evaluate the accuracy of prediction for eye, hair and skin pigmentation in adataset of > 6,500 individuals from Mexico, Colombia, Peru, Chile and Brazil (includinggenome-wide SNP data and quantitative/categorical pigmentation phenotypes - theCANDELA dataset CAN). We evaluated accuracy in relation to different analyticalmethods and various phenotypic predictors. As expected from statistical principles, weobserve that quantitative traits are more sensitive to changes in the prediction modelsthan categorical traits. We find that Random Forest or Linear Regression are generallythe best performing methods. We also compare the prediction accuracy of SNP setsdefined in the CAN dataset (including 56, 101 and 120 SNPs for eye, hair and skincolour prediction, respectively) to the well-established HIrisPlex-S SNP set (including 6,22 and 36 SNPs for eye, hair and skin colour prediction respectively). When trainingprediction models on the CAN data, we observe remarkably similar performances forHIrisPlex-S and the larger CAN SNP sets for the prediction of hair (categorical) andeye (both categorical and quantitative), while the CAN sets outperform HIrisPlex-S forquantitative, but not for categorical skin pigmentation prediction. The performance ofHIrisPlex-S, when models are trained in a world-wide sample (although consisting of80% Europeans, https://hirisplex.erasmusmc.nl), is lower relative to training in theCAN data (particularly for hair and skin colour). Altogether, our observations areconsistent with common variation of eye and hair colour having a relatively simplegenetic architecture, which is well captured by HIrisPlex-S, even in admixed LatinAmericans (with partial European ancestry). By contrast, since skin pigmentation is amore polygenic trait, accuracy is more sensitive to prediction SNP set size, althoughhere this effect was only apparent for a quantitative measure of skin pigmentation. Ourresults support the use of HIrisPlex-S in the prediction of categorical pigmentationtraits for forensic purposes in Latin America, while illustrating the impact of trainingdatasets on its accuracy.