Instability of high polygenic risk classification and mitigation by integrative scoring

Area of Science:

• Genetics and Genomics
• Bioinformatics
• Personalized Medicine

Background:

• Polygenic risk scores (PRS) are increasingly used in healthcare and commercial settings.
• The classification of high polygenic risk can vary significantly across different PRS.
• Novel methods and expanding genome-wide association studies (GWAS) continue to refine PRS.

Purpose of the Study:

• To assess the association and classification performance of cataloged PRS for three complex traits.
• To evaluate the consistency of high-risk individual classification across different PRS.
• To introduce and validate the PRSmix framework (ChronoAdd) for improved PRS integration and classification stability.

Main Methods:

• Chronological ordering of trait-related publications (Pub_n) and identification of the best-performing PRS (Best(Pub_n)) for each publication.
• Analysis of population-level association strengths and classification performance (top 10%) for individual PRS.
• Generation of ChronoAdd(Pub_n) scores using the PRSmix framework, integrating all PRS from publications up to Pub_n.

Main Results:

• Individual Best(Pub_n) PRS showed consistent population-level association strengths but wide variation in classifying high-risk individuals.
• ChronoAdd(Pub_n) scores demonstrated more consistent high-risk classification compared to individual Best(Pub_n) scores.
• The PRSmix framework provides a stable and reliable method for identifying high-risk individuals.

Conclusions:

• An integrative scoring approach using PRSmix offers stable and reliable classification of high-risk individuals.
• The ChronoAdd framework is adaptable for incorporating new PRS as they become available.
• This integrative strategy facilitates easier integration with current PRS implementation in clinical and commercial settings.