Refining models of archaic admixture in Eurasia with ArchaicSeeker 2.0

Kai Yuan,Jiaojiao Liu,Xueling Ge,Xixian Ma,Yuwen Pan,Xumin Ni,Haiyi Lou,Taoyang Wu,Lian Deng,Chang Liu,Rui Zhang,Shuhua Xu,Yang Gao

doi:10.1038/s41467-021-26503-5

Abstract

We developed a method, ArchaicSeeker 2.0, to identify introgressed hominin sequences and model multiple-wave admixture. The new method enabled us to discern two waves of introgression from both Denisovan-like and Neanderthal-like hominins in present-day Eurasian populations and an ancient Siberian individual. We estimated that an early Denisovan-like introgression occurred in Eurasia around 118.8–94.0 thousand years ago (kya). In contrast, we detected only one single episode of Denisovan-like admixture in indigenous peoples eastern to the Wallace-Line. Modeling ancient admixtures suggested an early dispersal of modern humans throughout Asia before the Toba volcanic super-eruption 74 kya, predating the initial peopling of Asia as proposed by the traditional Out-of-Africa model. Survived archaic sequences are involved in various phenotypes including immune and body mass (e.g., ZNF169), cardiovascular and lung function (e.g., HHAT), UV response and carbohydrate metabolism (e.g., HYAL1/HYAL2/HYAL3), while “archaic deserts” are enriched with genes associated with skin development and keratinization.

Highlights

IntroductionArchaicSeeker 2.0, to identify introgressed hominin sequences and model multiple-wave admixture
We developed a method, ArchaicSeeker 2.0, to identify introgressed hominin sequences and model multiple-wave admixture
We propose a generalized method called ArchaicSeeker 2.0 to simultaneously detect ancient sequences in the modern human genomes that are derived from archaic hominins and infer the introgression history

Summary

Introduction

ArchaicSeeker 2.0, to identify introgressed hominin sequences and model multiple-wave admixture. One common limitation of them is the difficulty to determine the precise boundaries of introgression segments To address this shortcoming, several methods are proposed under a unifying postulate that genomic regions with introgressed sequences should have distinct patterns compared with non-introgressed regions. These methods are generally based on sequential pattern recognition models such as the Hidden Markov Model (HMM)[2,18,23,25,27] and Conditional Random Field[17] While most of these methods detect introgressed sequences by comparing the test introgressed genome with archaic and African genomes, Hu et al.[25] and Skov et al.[18] used the information of non-AMH marker density to identify the introgressed sequences

Methods

Results

Conclusion