Evaluation of Polygenic Prediction Methodology within a Reference-Standardized Framework
Authors:
- Oliver PainOPOliver Pain
- Kylie Patricia GlanvilleKGKylie Patricia Glanville
- Saskia P. HagenaarsSHSaskia P. Hagenaars
- Saskia P SelzamSSSaskia P Selzam
- Anna E. FürtjesAFAnna E. Fürtjes
- Héléna Alexandra GasparHGHéléna Alexandra Gaspar
- Jonathan R. I. ColemanJonathan R. I. Coleman
- Kaili RimfeldKRKaili Rimfeld
- Gerome BreenGBGerome Breen
- Robert PlominRPRobert Plomin
- Lasse FolkersenLFLasse Folkersen
- Cathryn M. LewisCLCathryn M. Lewis
Preprint
Abstract
Background: The predictive utility of polygenic scores is increasing, and many polygenic scoring methods are available, but it is unclear which method performs best. This study evaluates the predictive utility of polygenic scoring methods within a reference-standardized framework, which uses a common set of variants and reference-based estimates of linkage disequilibrium and allele frequencies to construct scores. Methods: Eight polygenic score methods were tested: p-value thresholding and clumping (pT+clump), SBLUP, lassosum, LDPred1, LDPred2, PRScs, DBSLMM and SBayesR, evaluating their performance to predict outcomes in UK Biobank and the Twins Early Development Study (TEDS). Strategies to identify optimal p-value threshold and shrinkage parameters were compared, including 10-fold cross validation, pseudovalidation and infinitesimal models (with no validation sample), and multi-polygenic score elastic net models. Results: LDPred2, lassosum and PRScs performed strongly using 10-fold cross-validation to identify the most predictive p-value threshold or shrinkage parameter, giving a relative improvement of 16-18% over pT+clump in the correlation between observed and predicted outcome values. Using pseudovalidation, the best methods were PRScs and DBSLMM, with a relative improvement of >10% over other pseudovalidation and infinitesimal methods (lassosum, SBLUP, SBayesR, LDPred1, LDPred2). PRScs pseudovalidation was only 3% worse than the best polygenic score identified by 10-fold cross validation. Elastic net models containing polygenic scores based on a range of parameters consistently improved prediction over any single polygenic score. Conclusion: Within a reference-standardized framework, the best polygenic prediction was achieved using LDPred2, lassosum and PRScs, modeling multiple polygenic scores derived using multiple parameters. This study will help researchers performing polygenic score studies to select the most powerful and predictive analysis methods.Keywords
All Related Versions
- Published version: PLOS Genetics, 17 (5), e1009021.
This publication has 41 references indexed in Scilit:
- Explicit Modeling of Ancestry Improves Polygenic Risk Scores and BLUP PredictionGenetic Epidemiology, 2015
- Meta-analysis of the heritability of human traits based on fifty years of twin studiesNature Genetics, 2015
- Correction: Power and Predictive Accuracy of Polygenic Risk ScoresPLOS Genetics, 2013
- Power and Predictive Accuracy of Polygenic Risk ScoresPLOS Genetics, 2013
- A Better Coefficient of Determination for Genetic Profile AnalysisGenetic Epidemiology, 2012
- Integrating common and rare genetic variation in diverse human populationsNature, 2010
- Finding the missing heritability of complex diseasesNature, 2009
- Common polygenic variation contributes to risk of schizophrenia and bipolar disorderNature, 2009
- Regularization and Variable Selection Via the Elastic NetJournal of the Royal Statistical Society Series B: Statistical Methodology, 2005
- Tests for comparing elements of a correlation matrix.Psychological Bulletin, 1980