AncestryGeni: a novel genetic ancestry classification pipeline for small and noisy sequence data

Area of Science:

• Genomics
• Computational Biology
• Population Genetics

Background:

• Addressing health disparities requires accurate genetic ancestry inference.
• Existing computational tools face limitations with real-world genomic data, such as limited DNA and small marker sets.
• Self-described race is an inadequate proxy for genetic ancestry.

Purpose of the Study:

• To develop a novel supervised machine-learning tool, AncestryGeni, for robust genetic ancestry inference.
• To enable accurate ancestry estimation using limited genomic data, including whole exome sequencing (WES) and RNA sequencing (RNA-Seq).
• To overcome challenges posed by nonstandard genomic material and small sample sizes.

Main Methods:

• Developed AncestryGeni, a supervised machine-learning pipeline.
• Applied AncestryGeni to a real-world genomic dataset from the MMRF CoMMpass study.
• Validated AncestryGeni's accuracy against FastNGSadmix using nonstandard genomic material.

Main Results:

• AncestryGeni accurately infers genetic ancestry with as few as 100 markers.
• The tool demonstrates superior accuracy compared to FastNGSadmix on nonstandard genomic material.
• Tumor-derived sequences from WES and RNA-Seq are viable data sources for ancestry estimation with AncestryGeni.

Conclusions:

• AncestryGeni provides a robust and accurate method for genetic ancestry inference.
• The tool is applicable to diverse genomic data types, including WES and RNA-Seq.
• AncestryGeni facilitates research into health disparities by improving genetic ancestry determination.