Genotype error due to low-coverage sequencing induces uncertainty in polygenic scoring.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Polygenic scores (PGSs) are crucial for predicting phenotypes from genotype data.
• Traditional PGS methods often overlook genotyping errors from sequencing or imputation.
• These errors can affect the reliability of PGS estimations.

Purpose of the Study:

• To investigate the impact of genotyping errors from low-coverage whole-genome sequencing (lcWGS) on PGS accuracy.
• To develop and evaluate a probabilistic approach for PGS estimation that incorporates genotype error.
• To assess the implications of genotyping errors on PGS-based risk stratification.

Main Methods:

• Utilized SNP array and lcWGS data from 802 individuals (Dana-Farber PROFILE cohort).
• Analyzed the correlation between PGS error and sequencing depth.
• Developed a probabilistic PGS model accounting for genotype uncertainty.
• Employed simulations to explore combined error effects.

Main Results:

• PGS error significantly correlates with sequencing depth (p = 1.2 × 10⁻⁷).
• The probabilistic approach yielded well-calibrated PGS credible intervals.
• Classification accuracy improved by up to 6% compared to traditional PGS methods.
• Genotyping and effect size errors jointly impact risk prediction.

Conclusions:

• Genotyping error is a critical factor influencing PGS accuracy, particularly in lcWGS data.
• Probabilistic PGS modeling offers improved calibration and accuracy.
• Accounting for genotyping errors is essential for reliable PGS application in diverse cohorts.