Evaluating genetic-ancestry inference from single-cell RNA-seq data

Area of Science:

• Genomics
• Bioinformatics
• Population Genetics

Background:

• Characterizing donor ancestry in single-cell RNA sequencing (scRNA-seq) is vital for dataset homogeneity, bias reduction, and identifying ancestry-specific regulatory mechanisms.
• Missing donor ancestry information in scRNA-seq studies impedes comprehensive analysis and understanding of disease relevance.
• Ensuring scRNA-seq datasets represent global human genetic diversity is essential for equitable research outcomes.

Purpose of the Study:

• To propose and evaluate a framework for assessing methods to infer genetic ancestry from genetic polymorphisms detected in scRNA-seq data.
• To demonstrate the accuracy of established tools like ADMIXTURE for ancestry inference using scRNA-seq data, even with limited polymorphisms and imperfect variant calling.
• To infer genetic ancestry for donors in existing scRNA-seq datasets and highlight potential biases in population representation.

Main Methods:

• Developed a framework to evaluate genetic ancestry inference methods using genetic polymorphisms from scRNA-seq reads.
• Applied widely used bioinformatics tools, such as ADMIXTURE, to analyze scRNA-seq data for ancestry estimation.
• Inferred genetic ancestry for 196 donors across four Human Cell Atlas scRNA-seq datasets.

Main Results:

• Widely used tools accurately infer genetic ancestry and admixture proportions from scRNA-seq data.
• Inference remains robust despite limited genetic polymorphisms and imperfect variant calling inherent in scRNA-seq.
• Analysis of Human Cell Atlas datasets revealed a disproportionately high number of donors of European ancestry.

Conclusions:

• Genetic ancestry inference from scRNA-seq data is feasible and accurate using existing computational tools.
• Current scRNA-seq datasets, including those from the Human Cell Atlas, are heavily skewed towards European ancestry.
• Researchers are urged to report genetic ancestry for all donors and prioritize generating more diverse scRNA-seq datasets to enhance representation and reduce bias.