SALAI-Net: species-agnostic local ancestry inference network

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Local ancestry inference (LAI) predicts ancestry labels along DNA sequences, crucial for human history studies and precision medicine.
• Current LAI models struggle with generalization across species, chromosomes, or ancestry groups, necessitating re-training and often lacking interpretability.

Purpose of the Study:

• To develop a portable, species-agnostic local ancestry inference method that generalizes across diverse datasets and requires minimal biological parameters.
• To improve the accuracy, speed, and interpretability of LAI compared to existing approaches.

Main Methods:

• Introduced SALAI-Net, a statistical LAI method inspired by identity by descent principles.
• Employed a reference matching approach for estimating population labels on DNA segments, ensuring interpretability and speed.
• Validated models on human whole-genome data and tested cross-species generalization to canids.

Main Results:

• SALAI-Net demonstrates superior balanced accuracy compared to existing methods.
• The model exhibits strong generalization capabilities across different settings, species, and datasets.
• SALAI-Net is significantly faster (up to two orders of magnitude) and more memory-efficient than competing methods.

Conclusions:

• SALAI-Net provides a highly accurate, interpretable, and efficient solution for local ancestry inference.
• Its species-agnostic nature and strong generalization performance make it a versatile tool for various genetic applications.
• The open-source implementation facilitates broader adoption and further research in the field.