Polygenic risk scores: a biased prediction?

Area of Science:

• Genetics
• Bioinformatics
• Public Health

Background:

• Polygenic risk scores (PRS) are increasingly used to predict individual disease susceptibility.
• Current PRS models are primarily derived from European ancestry populations, leading to potential inaccuracies in other groups.
• Genotyping arrays, common research tools, introduce design biases that affect PRS performance across diverse ancestries.

Discussion:

• This study reveals significant biases and inaccuracies in polygenic risk score (PRS) predictions when applied to populations not included in the original derivation datasets.
• The inherent design biases of widely used research tools, such as genotyping arrays, are identified as a major contributor to these PRS distortions.
• These inaccuracies can exacerbate existing health inequities, particularly for underrepresented ethnic and ancestral groups.

Key Insights:

• Polygenic risk scores (PRS) demonstrate poor predictive accuracy for disease risk in non-derived populations.
• Genotyping array design introduces systematic biases, compromising the generalizability of PRS.
• Failure to address these biases perpetuates health disparities and limits the clinical utility of PRS in diverse individuals.

Outlook:

• Future PRS development must prioritize the inclusion of diverse ancestral populations to improve accuracy and equity.
• Development of novel genotyping technologies and analytical methods is needed to mitigate design biases.
• Implementing bias reduction strategies and ensuring equitable data representation are crucial for the responsible advancement of genomic medicine.