Estimating Disorder Probability Based on Polygenic Prediction Using the BPC Approach.

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Polygenic Scores (PGSs) aggregate genetic predisposition for traits using Genome-Wide Association Study (GWAS) data.
• Existing Bayesian PGS methods offer improved prediction accuracy for continuous traits but lack calibration for binary disorder traits in ascertained samples.
• Accurate calibration of PGSs is crucial for estimating absolute individual disorder probabilities for clinical applications.

Purpose of the Study:

• To introduce and evaluate the Bayesian polygenic score Probability Conversion (BPC) approach for calibrating PGSs for binary disorder traits.
• To enable reliable computation of absolute disorder probabilities for individuals.

Main Methods:

• The BPC approach utilizes GWAS summary statistics, Bayesian PGS methods (e.g., PRScs, SBayesR), individual genotype data, and a prior disorder probability.
• It involves transforming the PGS to a liability scale, calculating PGS variances in cases and controls, and applying Bayes' Theorem.
• The method is practical as it does not require a separate tuning sample with both genotype and phenotype data.

Main Results:

• The BPC approach demonstrated well-calibrated results on extensive simulated and empirical data for nine disorders.
• Performance was consistently superior compared to another recently published calibration approach.
• The method effectively converts PGSs to accurate absolute disorder probabilities.

Conclusions:

• The Bayesian polygenic score Probability Conversion (BPC) approach provides a robust and practical solution for calibrating polygenic scores for binary disorder prediction.
• This method facilitates the reliable estimation of absolute individual disorder probabilities, paving the way for clinical implementation.
• BPC offers improved accuracy and calibration over existing methods for genetic risk prediction in disorders.