Assessing agreement between different polygenic risk scores in the UK Biobank

Clifton, Lei; Collister, Jennifer A; Liu, Xiaonan; Littlejohns, Thomas J.; Hunter, David J.

doi:10.1101/2022.02.09.22270719

Cited by 4 publications

(5 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our study joins earlier studies in their call for making the use and reporting of PGIs and their construction more transparent and standardised 17–19,41 and contributes to the set of recent studies highlighting the divergent predictive power of PGIs 20,42–44 . Pain et al compare a very extensive set of traits and test the predictive power of a wide variety of PGI construction methods 43 .…”

Section: Discussionsupporting

confidence: 64%

Rank concordance of polygenic indices: Implications for personalised intervention and gene-environment interplay

Muslimova

Pereira

Hinke

et al. 2022

Preprint

View full text Add to dashboard Cite

Polygenic indices (PGIs) are increasingly used to identify individuals at high risk of developing diseases and disorders and are advocated as a screening tool for personalised intervention in medicine and education. The performance of PGIs is typically assessed in terms of the amount of phenotypic variance they explain in independent prediction samples. However, the correct ranking of individuals in the PGI distribution is a more important performance metric when identifying individuals at high genetic risk. We empirically assess the rank concordance between PGIs that are created with different construction methods and discovery samples, focusing on cardiovascular disease (CVD) and educational attainment (EA). We find that the rank correlations between the constructed PGIs vary strongly (Spearman correlations between 0.17 and 0.94 for CVD, and between 0.40 and 0.85 for EA), indicating highly unstable rankings across different PGIs for the same trait. Simulations show that measurement error in PGIs is responsible for a substantial part of PGI rank discordance. Potential consequences for personalised medicine in CVD and research on gene-environment (G×E) interplay are illustrated using data from the UK Biobank.

show abstract

Section: Discussionsupporting

confidence: 64%

Rank concordance of polygenic indices: Implications for personalised intervention and gene-environment interplay

Muslimova

Pereira

Hinke

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Our study joins earlier studies in their call for making the use and reporting of PGIs and their construction more transparent and standardised [17][18][19]41 and contributes to the set of recent studies highlighting the divergent predictive power of PGIs 20,[42][43][44] . Pain et al compare a very extensive set of traits and test the predictive power of a wide variety of PGI construction methods 43 .…”

Section: Personalised Interventionssupporting

confidence: 63%

“…Ware et al compare a more limited set of PGI construction methods and analyse the intra-individual correlation of PGIs 42 . Finally, two studies that were independently developed around the same time 20,44 are similar in spirit as the present study in comparing the rank concordance of individuals in the PGI distribution depending on the GWAS discovery sample. Our study complement these studies by i) explicitly focusing on rank discordance and its source, ii) comparing across PGI construction methods (e.g., C+T and LDpred), and iii) analysing the implications for empirical applications such as personalised medicine or G×E analysis.…”

Section: Personalised Interventionsmentioning

confidence: 70%

Rank concordance of polygenic indices

et al. 2023

View full text Add to dashboard Cite

Polygenic indices (PGIs) are increasingly used to identify individuals at high risk of developing diseases and disorders and are advocated as a screening tool for personalised intervention in medicine and education. The performance of PGIs is typically assessed in terms of the amount of phenotypic variance they explain in independent prediction samples. However, the correct ranking of individuals in the PGI distribution is a more important performance metric when identifying individuals at high genetic risk.We empirically assess the rank concordance between PGIs that are created with different construction methods and discovery samples, focusing on cardiovascular disease (CVD) and educational attainment (EA). We find that the rank correlations between the constructed PGIs vary strongly (Spearman correlations between 0.17 and 0.94 for CVD, and between 0.40 and 0.85 for EA), indicating highly unstable rankings across different PGIs for the same trait. Simulations show that measurement error in PGIs is responsible for a substantial part of PGI rank discordance. Potential consequences for personalised medicine in CVD and research on gene-environment (G×E) interplay are illustrated using data from the UK Biobank.

show abstract

“…Ensembling of epistatic models with PRS did not give the same improvement as was observed in the cross-validation analysis, likely because the PRS model performed poorly on the holdout dataset in general. We suspect this is due to the well-known difficulties in applying PRS models trained on one dataset to another [36]. However, ensembling with a PRS model did improve model specificity and positive predictive value on the holdout set.…”

Section: Discussionmentioning

confidence: 94%

Epistatic Features and Machine Learning Improve Alzheimer’s Risk Prediction Over Polygenic Risk Scores

Hermes

Cady

Armentrout

et al. 2023

Preprint

View full text Add to dashboard Cite

Background. Polygenic risk scores (PRS) are linear combinations of genetic markers weighted by effect size that are commonly used to predict disease risk. For complex heritable diseases such as late onset Alzheimer's disease (LOAD), PRS models fail to capture much of the heritability. Additionally, PRS models are highly dependent on the population structure of data on which effect sizes are assessed, and have poor generalizability to new data. Objective. The goal of this study is to construct a paragenic risk score that, in addition to single genetic marker data used in PRS, incorporates epistatic interaction features and machine learning methods to predict lifetime risk for LOAD. Methods. We construct a new state-of-the-art genetic model for lifetime risk of Alzheimer's disease. Our approach innovates over PRS models in two ways: First, by directly incorporating epistatic interactions between SNP loci using an evolutionary algorithm guided by shared pathway information; and second, by estimating risk via an ensemble of machine learning models (gradient boosting machines and deep learning) instead of simple logistic regression. We compare the paragenic model to a PRS model from the literature trained on the same dataset. Results. The paragenic model is significantly more accurate than the PRS model under 10-fold cross-validation, obtaining an AUC of 83% and near-clinically significant matched sensitivity/specificity of 75%, and remains significantly more accurate when evaluated on an independent holdout dataset. Additionally, the paragenic model maintains accuracy within APOE genotypes. Conclusion. Paragenic models show potential for improving lifetime disease risk prediction for complex heritable diseases such as LOAD over PRS models.

show abstract

Assessing agreement between different polygenic risk scores in the UK Biobank

Cited by 4 publications

References 29 publications

Rank concordance of polygenic indices: Implications for personalised intervention and gene-environment interplay

Rank concordance of polygenic indices: Implications for personalised intervention and gene-environment interplay

Rank concordance of polygenic indices

Epistatic Features and Machine Learning Improve Alzheimer’s Risk Prediction Over Polygenic Risk Scores

Contact Info

Product

Resources

About