Joint modeling of genetically correlated diseases and functional annotations increases accuracy of polygenic risk prediction

Hu, Yiming; Lu, Qiongshi; Liu, Wei; Zhang, Yuhua; Li, Mo; Zhao, Hongyu

doi:10.1371/journal.pgen.1006836

Cited by 75 publications

(64 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the predictive performance of polygenic risk score (PRS) models are relatively poor for most of complex diseases, PRS will be improved by increasing the sample size of the training GWAS data set and innovative statistical methods that incorporate additional biological information, e.g., functional annotation data and genetic pleiotropy (Chen, 2018;Hu, et al, 2017;Shi, et al, 2016). One difficulty for developing more accurate PRS is to evaluate the predictive LD-Pred (Vilhjalmsson, et al, 2015) and BLUP-type PRSs (Golan and Rosset, 2014;Speed and Balding, 2014) that are based on linear mixed models.…”

Section: Discussionmentioning

confidence: 99%

SummaryAUC: a tool for evaluating the performance of polygenic risk prediction models in validation datasets with only summary level statistics

Song¹,

Liu

Shi³

2018

Preprint

View full text Add to dashboard Cite

MotivationPolygenic risk score (PRS) methods based on genome-wide association studies (GWAS) have a potential for predicting the risk of developing complex diseases and are expected to become more accurate with larger training data sets and innovative statistical methods. The area under the ROC curve (AUC) is often used to evaluate the performance of PRSs, which requires individual genotypic and phenotypic data in an independent GWAS validation dataset. We are motivated to develop methods for approximating AUC of PRSs based on the summary level data of the validation dataset, which will greatly facilitate the development of PRS models for complex diseases.ResultsWe develop statistical methods and an R package SummaryAUC for approximating the AUC and its variance of a PRS when only the summary level data of the validation dataset are available. SummaryAUC can be applied to PRSs with SNPs either genotyped or imputed in the validation dataset. We examined the performance of SummaryAUC using a large-scale GWAS of schizophrenia. SummaryAUC provides accurate approximations to AUCs and their variances. The bias of AUC is typically less than 0.5% in most analyses. SummaryAUC cannot be applied to PRSs that use all SNPs in the genome because it is computationally prohibitive.Availabilityhttps://github.com/lsncibb/SummaryAUCContactJianxin.Shi@nih.gov

show abstract

Section: Discussionmentioning

confidence: 99%

SummaryAUC: a tool for evaluating the performance of polygenic risk prediction models in validation datasets with only summary level statistics

Song¹,

Liu

Shi³

2018

Preprint

View full text Add to dashboard Cite

show abstract

“…Figure S9: The enrichment results of single-cell RNA sequencing (scRNAseq) data in different communities. These results are for five communities generated by GeNets 79 . For each community, scRNAseq data were tested for genes from gTADA only.…”

Section: Simulation Of Data To Test the False Discovery Rates Of Top mentioning

confidence: 99%

Integrative analysis of rare variants and pathway information shows convergent results between immune pathways, drug targets and epilepsy genes

Dobbyn

et al. 2018

Preprint

View full text Add to dashboard Cite

Trio family and case-control studies of next-generation sequencing data have proven integral to understanding the contribution of rare inherited and de novo single-nucleotide variants to the genetic architecture of complex disease. Ideally, such studies should identify individual risk genes of moderate to large effect size to generate novel treatment hypotheses for further follow-up. However, due to insufficient power, gene set enrichment analyses have come to be relied upon for detecting differences between cases and controls, implicating sets of hundreds of genes rather than specific targets for further investigation. Here, we present a Bayesian statistical framework, termed gTADA, that integrates gene-set membership information with gene-level de novo and rare inherited case-control counts, to prioritize risk genes with excess rare variant burden within enriched gene sets. Applying gTADA to available whole-exome sequencing datasets for several neuropsychiatric conditions, we replicated previously reported gene set enrichments and identified novel risk genes. For epilepsy, gTADA prioritized 40 risk genes (posterior probabilities > 0.95), 6 of which replicate in an independent whole-genome sequencing study. In addition, 30/40 genes are novel genes. We found that epilepsy genes had high protein-protein interaction (PPI) network connectivity, and show specific expression during human brain development. Some of the top prioritized EPI genes were connected to a PPI subnetwork of immune genes and show specific expression in prenatal microglia. We also identified multiple enriched drug-target gene sets for EPI which included immunostimulants as well as known antiepileptics. Immune biology was supported specifically by case-control variants from familial epilepsies rather than do novo mutations in generalized encephalitic epilepsy. meta-analyzing DNMs and rare case-control (CC) variants, an approach that has been particularly successful for autism spectrum disorders (ASD) 9,10 . For epilepsy (EPI), multiple associated genes have been identified through DN based studies 4,5,11 , and in recent years, a number of EPI significant genes have also been identified through CC studies 12,13 . We hypothesized that, as for ASD, additional significant EPI genes could be discovered through the integration of DN and CC data. EPI is a serious brain disorder which includes multiple subtypes. Studies of cases/controls and twins have shown that genetic components have played important roles in EPI [14][15][16] . Some of EPI's subtypes can be explained by single genes, but multiple subtypes might be caused by multiple genes 15 . It is still challenging to develop specific drugs for this disorder. There have been multiple antiepileptic drugs used for EPI treatments; however, 20-30% of EPI patients have not been successful in controlling their seizures by using current medications 17 . Identifying additional genes or gene sets might help better understand its etiology as well as better design drug targets for the disorder.Due to the high p...

show abstract

“…Prediction accuracy of PRS remains moderate for most phenotypes. 13 Methods have been developed to improve PRS performance by explicitly modeling linkage disequilibrium (LD), 14 incorporating functional annotations and pleiotropy, 15,16 and improving effect estimates through statistical shrinkage. 17 Notably, most PRS models have tuning parameters, including the p-value threshold in traditional PRS, the penalty strength in penalized regression models, and the proportion of causal variants in LDpred.…”

Section: Introductionmentioning

confidence: 99%

Fine-tuning Polygenic Risk Scores with GWAS Summary Statistics

Zhao

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Polygenic risk scores (PRSs) have wide applications in human genetics research. Notably, most PRS models include tuning parameters which improve predictive performance when properly selected. However, existing model-tuning methods require validation data that is independent with both training and testing samples. These data rarely exist in practice, creating a significant gap between PRS methodology and applications. Here, we introduce PUMAS, a novel method to finetune PRS models using summary statistics from genome-wide association studies (GWASs). Through extensive simulations, external validations, and analysis of 65 GWAS traits, we demonstrate that PUMAS can perform a variety of model-tuning procedures (e.g. cross-validation) using GWAS summary statistics and can effectively benchmark and optimize PRS models under diverse genetic architecture. Applied to 211 neuroimaging traits and Alzheimer's disease, we show that fine-tuned PRSs will improve statistical power in association analysis. We believe our method resolves a fundamental problem without a current solution and will greatly benefit genetic prediction applications.

show abstract

Joint modeling of genetically correlated diseases and functional annotations increases accuracy of polygenic risk prediction

Cited by 75 publications

References 36 publications

SummaryAUC: a tool for evaluating the performance of polygenic risk prediction models in validation datasets with only summary level statistics

SummaryAUC: a tool for evaluating the performance of polygenic risk prediction models in validation datasets with only summary level statistics

Integrative analysis of rare variants and pathway information shows convergent results between immune pathways, drug targets and epilepsy genes

Fine-tuning Polygenic Risk Scores with GWAS Summary Statistics

Contact Info

Product

Resources

About