ArchIE2: A software package for robust inference of introgressed local ancestry

Human Population Genetics and Genomics 2026;6(1):0002 | https://doi.org/10.47248/hpgg2606010002

Original Research Open Access

ArchIE2: A software package for robust inference of introgressed local ancestry

Harold Wang ¹ , Sriram Sankararaman ^2,3,4

1 Bioinformatics Interdepartmental Program, UCLA, Los Angeles, CA 90095, USA
2 Department of Computer Science, UCLA, Los Angeles, CA 90095, USA
3 Department of Computational Medicine, David Geffen School of Medicine, UCLA, Los Angeles, CA 90095, USA
4 Department of Human Genetics, David Geffen School of Medicine, UCLA, Los Angeles, CA 90095, USA

Correspondence: Harold Wang; Sriram Sankararaman

Academic Editor(s): Joshua Akey, Carina Schlebusch, Torsten Günther

Received: Sep 18, 2025 | Accepted: Dec 13, 2025 | Published: Jan 17, 2026

This article belongs to the Special Issue Population Genetics Methods and Software

Cite this article: Wang H, Sankararaman S. ArchIE2: A software package for robust inference of introgressed local ancestry. Hum Popul Genet Genom. 2026;6(1):0002. https://doi.org/10.47248/hpgg2606010002

Abstract

Introgression is a pervasive feature of human and non-human evolutionary history, and methods that identify introgressed loci have become central to studying its biological impact. We present ArchIE2, an enhanced and more robust version of the reference-free local ancestry framework introduced by ArchIE. ArchIE2 replaces ArchIE’s high-dimensional, sample-size–dependent feature set with a compact collection of summary statistics that generalize across demographic settings. This redesign removes dependency to sample size while preserving predictive accuracy and improving model stability. Across simulations, ArchIE2 matches or exceeds the performance of existing approaches, demonstrating a flexible and scalable framework for detecting introgressed segments, including in scenarios lacking reference genomes from source populations.

Keywords

introgression, admixture, population genetics, machine learning, human evolution

Full-Text PDF

Full-Text HTML

Share this article