Recomb-Mix: fast and accurate local ancestry inference

Area of Science:

• Genetics
• Population Genetics
• Bioinformatics

Background:

• Large genotyped cohorts enable high-resolution genetic structure analysis of admixed populations using local ancestry inference (LAI).
• Current LAI methods face challenges with similar reference populations, numerous references, or deep admixture events, common in large biobanks.

Purpose of the Study:

• To introduce Recomb-Mix, a novel LAI method designed to overcome limitations of existing approaches.
• To improve the accuracy and efficiency of LAI in complex admixed population scenarios.

Main Methods:

• Recomb-Mix integrates elements of the site-based Li and Stephens model.
• Novel graph collapsing techniques are employed to simplify path counting with identical ancestry labels.
• The method was benchmarked on simulated and real-world datasets.

Main Results:

• Recomb-Mix demonstrates superior accuracy compared to existing LAI methods across diverse simulated scenarios.
• The method shows competitive resource efficiency.
• Scalability and robustness were validated using real-world population data.

Conclusions:

• Recomb-Mix provides a more accurate and efficient solution for LAI in admixed populations.
• The method is expected to advance genetic studies of populations with complex admixture histories.
• Recomb-Mix is available as an open-source implementation.