Comparing Neanderthal Introgression Maps Reveals Core Agreement But Substantial Heterogeneity.

Area of Science:

• Genomics
• Population Genetics
• Paleogenetics

Background:

• Statistical methods for identifying Neanderthal ancestry in modern human genomes rely on diverse assumptions and inputs.
• Most studies utilize a single method, limiting understanding of comparative accuracy and sensitivity.

Purpose of the Study:

• To perform a large-scale comparison of 14 genome-wide introgression maps generated by 11 distinct Neanderthal introgression detection algorithms.
• To assess the accuracy, strengths, weaknesses, and downstream sensitivity of various introgression detection methods.

Main Methods:

• Compared 14 genome-wide introgression maps from 11 algorithms (admixfrog, ArchaicSeeker2, ArchIE, ARGWeaver-D, CRF, DICAL-ADMIX, hmmix, IBDmix, SARGE, Sprime, S*).
• Algorithms employed diverse statistical approaches including summary statistics, probabilistic modeling, and machine learning.
• Input data varied across algorithms, utilizing archaic, modern, and simulated genomes.

Main Results:

• Identified a core set of Neanderthal introgression regions consistently predicted across most methods.
• Observed substantial heterogeneity in introgression maps, particularly at the individual genome level.
• Downstream analyses yielded different conclusions based on the specific introgression map employed.

Conclusions:

• Recommends careful selection of introgression maps for downstream analyses.
• Advocates for the use of multiple Neanderthal introgression maps to ensure the robustness of research findings.
• Provides integrated prediction sets to facilitate further study of Neanderthal introgression's impact on modern human populations.