![]() ![]() Although the iPSYCH cohort has been designed around psychiatric disorders, the study individuals can be linked to the National Danish Registers 22, 23, making it possible to generate multi-PGS for any phenotype captured in these registers. We benchmark the multi-PGS against each phenotype’s respective single PGS prediction and compare it with an existing PGS method that meta-analyzes multiple PGSs using GWAS summary statistics, wMT-SBLUP 8. These disorders are genetically correlated with many other psychiatric and neurological disorders as well as other behavioral phenotypes 24, 25, which are precisely the circumstances under which the proposed multi-PGS might boost the polygenic prediction accuracy. ![]() We apply our multi-PGS framework to the Lundbeck Foundation Initiative for Integrative Psychiatric Research (iPSYCH) 22, 23, one of the largest datasets on the genetics of major psychiatric disorders. This model is then evaluated in an independent dataset in terms of the prediction accuracy of the multi-PGS. Multiple PGS and covariates can be combined using either a linear model (lasso penalized regression) or a nonlinear model (XGBoost 21) into a multi-PGS model. ![]() These multi-PGS models can be trained on thousands of different PGS such as for health outcomes, body measurements, and behavioral phenotypes which are not necessarily genetically correlated with the outcome. In this work, we propose and evaluate a multi-PGS framework that leverages these key PGS developments to construct more powerful and generalisable prediction models. Therefore, as only one PGS per phenotype needs to be included in the prediction model, the only practical limitation is effectively the number of individual-level samples available for the desired phenotype, which is constantly growing for biobank data. Second, PGS for any genetically correlated phenotype, even those not available in the target data, can now be included easily in the same prediction model, which significantly expands the set of phenotypes one can study. First, individual-level genotype validation data for each of the correlated phenotypes included in the prediction model is no longer necessary because selecting the best-performing hyper-parameters is no longer needed. This development has two major implications in the context of prediction using multiple PGS. ![]() Recent advances in PGS methods allow us to generate scores for a phenotype without requiring validation data to tune the hyper-parameters 16, 17, 18, 19, 20. 10 for each GWAS summary statistic number), which increases the number of validation samples required to fit the model and limits the number of PGS that can be included in practice. For the latter, the number of included PGS can become very large (e.g. Previously proposed prediction methods using multiple PGS have either required individual-level genotype and phenotype validation data for each PGS in the model 7, 11, or the inclusion of multiple PGS for the same GWAS summary statistics file corresponding to different p-value thresholds or proportions of causal variants 15 (i.e. Increasing the number of samples for a phenotype is costly and takes time, but a possible alternative is to use genetically correlated phenotypes to increase the effective sample size at no cost 1, 7, 8, 9, 10, 11, 12, 13, 14. The predictive performance of PGS is largely determined by four factors: the sample size of the GWAS used for training the score, the proportion of causal variants and the heritability of the phenotype, as well as heterogeneity between GWAS and target samples, including differences in genetic ancestry 4, 5, 6. Nature Communications volume 14, Article number: 4702 ( 2023)Īlthough polygenic scores (PGS) have high potential for clinical use 1, 2, 3, they are currently underpowered for many applications regarding disease prediction and risk stratification. Multi-PGS enhances polygenic prediction by combining 937 polygenic scores ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |