Hide/Show Apps

Benchmarking kinship estimation tools for ancient genomes using pedigree simulations

2020-10-22
Çetin, Mehmet
Aktürk, Şevval
Mapelli, Igor
Yaka, Reyhan
Çokoğlu, Seda
Zakaria, Douaa
Ceballos, Francisco
Ghalichi, Ayshin
Altınışık, N. Ezgi
Koptekin, Dilek
Vural, Kıvılcım Başak
Erdal, Yılmaz Selim
Atakuman, Çiğdem
Götherström, Anders
Özer, Füsun
Sürer, Elif
Somel, Mehmet
There is growing interest in identifying genetic kinship levels among ancient individuals buried in physical proximity, from graveyards to archaeological crime scenes. Ancient genome data produced using massively parallel sequencing allow such estimation, albeit with significant limitations owing to the extremely low coverage of ancient genomes, frequently ranging between 0.2x - 0.01x. Estimating the kinship coefficient reliably using such sparse data is a challenge. Both likelihood-based and non-parametric tools have been recently developed to address this question, but their efficiency has not yet been systematically and comparatively studied. Here we present work where we compared the accuracy of three most commonly employed tools, using ancient genome data produced from simulated pedigrees. We studied accuracy with respect to both kinship coefficient estimation, and also Cotterman coefficients. Our results show that genotype data that include >5,000 SNPs allow close kinship coefficients to be relatively reliably estimated, although accuracy falls dramatically beyond the 3rd degree. In addition, we confirm that pedigree relationship estimation (e.g. distinguishing between parent-child pairs vs. siblings) using Cotterman coefficients is a noteworthy problem, and will require alternative, and holistic approaches to address.