An optimized method to infer relatedness up to the 5th degree from low coverage ancient human genomes
BackgroundCurrent state of art kinship analysis is capable to infer relatedness up to the 5-6th degree from deeply sequenced DNA if the proper reference population is known. Low coverage, partially genotyped, degraded archaic (or forensic) DNA and often unavailable or unknown reference population poses additional challenges, hence kinship analysis from low coverage archaic sequences so far has been possible up to the second degree with large uncertainties. ResultsWe performed extensive simulations to identify and correct the main factors of bias in kinship analysis from low coverage data. As a result, we introduce a new metric for correction and offer a guideline, which overcomes the difficulties associated with low coverage samples. We validated our methodology on experimental modern and archaic data with widely different genome coverages (0.12x-11.9x) using samples with known family relations and known or unknown population structure. Out of 2526 ancient individuals from the REICH data set we confirmed all 96 indicated, and identified 303 new relatives additionally. ConclusionWith the proposed workflow we provide the necessary additional tools to calculate the corrected kinship coefficient from the commonly used genome data formats. Our methodology allows to reliably identify relatedness up to the 4-5th degree from variable/low coverage archaic (or badly degraded forensic) WGS genome data.