Ancestry inference using reference labeled clusters of haplotypes.

Area of Science:

• Population Genetics
• Genomic Ancestry Inference
• Computational Biology

Background:

• Accurate inference of human ancestry composition is crucial for genetic studies.
• Existing methods may struggle with complex population admixture.
• Haplotype-based approaches offer a promising avenue for ancestry analysis.

Purpose of the Study:

• To introduce ARCHes, a novel, fast, and accurate haplotype-based method for inferring individual ancestry composition.
• To evaluate the performance of ARCHes against established methods, particularly in admixed populations.

Main Methods:

• ARCHes models haplotype diversity from large admixed cohorts.
• Population information from reference panels is used to annotate these models.
• Training and testing are separate, allowing for efficient reuse of models on large datasets.

Main Results:

• ARCHes demonstrates fast computational performance, averaging under one minute per individual for 32 populations on 10 CPUs.
• The method was validated on diverse datasets, including the 1000 Genomes Project and HGDP.
• ARCHes significantly outperforms RFMix in correctly assigning both global and local ancestry.

Conclusions:

• ARCHes provides a superior approach for ancestry inference, especially in populations with complex admixture.
• The speed and accuracy of ARCHes make it a valuable tool for large-scale genomic studies.
• This method advances the field of population genetics by enabling more precise ancestry deconvolution.