Polygenic risk score portability for common diseases across genetically diverse populations.

BACKGROUND: Polygenic risk scores (PRS) derived from European individuals have reduced portability across global populations, limiting their clinical implementation at worldwide scale. Here, we investigate the performance of a wide range of PRS models across four ancestry groups (Africans, Europeans, East Asians, and South Asians) for 14 conditions of high-medical interest. METHODS: To select the best-performing model per trait, we first compared PRS performances for publicly available scores, and constructed new models using different methods (LDpred2, PRS-CSx and SNPnet). We used 285 K European individuals from the UK Biobank (UKBB) for training and 18 K, including diverse ancestries, for testing. We then evaluated PRS portability for the best models in Europeans and compared their accuracies with respect to the best PRS per ancestry. Finally, we validated the selected PRS models using an independent set of 8,417 individuals from Biobank of the Americas-Genomelink (BbofA-GL); and performed a PRS-Phewas. RESULTS: We confirmed a decay in PRS performances relative to Europeans when the evaluation was conducted using the best-PRS model for Europeans (51.3% for South Asians, 46.6% for East Asians and 39.4% for Africans). We observed an improvement in the PRS performances when specifically selecting ancestry specific PRS models (phenotype variance increase: 1.62 for Africans, 1.40 for South Asians and 0.96 for East Asians). Additionally, when we selected the optimal model conditional on ancestry for CAD, HDL-C and LDL-C, hypertension, hypothyroidism and T2D, PRS performance for studied populations was more comparable to what was observed in Europeans. Finally, we were able to independently validate tested models for Europeans, and conducted a PRS-Phewas, identifying cross-trait interplay between cardiometabolic conditions, and between immune-mediated components. CONCLUSION: Our work comprehensively evaluated PRS accuracy across a wide range of phenotypes, reducing the uncertainty with respect to which PRS model to choose and in which ancestry group. This evaluation has let us identify specific conditions where implementing risk-prioritization strategies could have practical utility across diverse ancestral groups, contributing to democratizing the implementation of PRS.

Validating and automating learning of cardiometabolic polygenic risk scores from direct-to-consumer genetic and phenotypic data: implications for scaling precision health research.

INTRODUCTION: A major challenge to enabling precision health at a global scale is the bias between those who enroll in state sponsored genomic research and those suffering from chronic disease. More than 30 million people have been genotyped by direct-to-consumer (DTC) companies such as 23andMe, Ancestry DNA, and MyHeritage, providing a potential mechanism for democratizing access to medical interventions and thus catalyzing improvements in patient outcomes as the cost of data acquisition drops. However, much of these data are sequestered in the initial provider network, without the ability for the scientific community to either access or validate. Here, we present a novel geno-pheno platform that integrates heterogeneous data sources and applies learnings to common chronic disease conditions including Type 2 diabetes (T2D) and hypertension. METHODS: We collected genotyped data from a novel DTC platform where participants upload their genotype data files and were invited to answer general health questionnaires regarding cardiometabolic traits over a period of 6 months. Quality control, imputation, and genome-wide association studies were performed on this dataset, and polygenic risk scores were built in a case-control setting using the BASIL algorithm. RESULTS: We collected data on N = 4,550 (389 cases / 4,161 controls) who reported being affected or previously affected for T2D and N = 4,528 (1,027 cases / 3,501 controls) for hypertension. We identified 164 out of 272 variants showing identical effect direction to previously reported genome-significant findings in Europeans. Performance metric of the PRS models was AUC = 0.68, which is comparable to previously published PRS models obtained with larger datasets including clinical biomarkers. DISCUSSION: DTC platforms have the potential of inverting research models of genome sequencing and phenotypic data acquisition. Quality control (QC) mechanisms proved to successfully enable traditional GWAS and PRS analyses. The direct participation of individuals has shown the potential to generate rich datasets enabling the creation of PRS cardiometabolic models. More importantly, federated learning of PRS from reuse of DTC data provides a mechanism for scaling precision health care delivery beyond the small number of countries who can afford to finance these efforts directly. CONCLUSIONS: The genetics of T2D and hypertension have been studied extensively in controlled datasets, and various polygenic risk scores (PRS) have been developed. We developed predictive tools for both phenotypes trained with heterogeneous genotypic and phenotypic data generated outside of the clinical environment and show that our methods can recapitulate prior findings with fidelity. From these observations, we conclude that it is possible to leverage DTC genetic repositories to identify individuals at risk of debilitating diseases based on their unique genetic landscape so that informed, timely clinical interventions can be incorporated.

Dynamic Feature Acquisition Using Denoising Autoencoders.

In real-world scenarios, different features have different acquisition costs at test time which necessitates cost-aware methods to optimize the cost and performance tradeoff. This paper introduces a novel and scalable approach for cost-aware feature acquisition at test time. The method incrementally asks for features based on the available context that are known feature values. The proposed method is based on sensitivity analysis in neural networks and density estimation using denoising autoencoders with binary representation layers. In the proposed architecture, a denoising autoencoder is used to handle unknown features (i.e., features that are yet to be acquired), and the sensitivity of predictions with respect to each unknown feature is used as a context-dependent measure of informativeness. We evaluated the proposed method on eight different real-world data sets as well as one synthesized data set and compared its performance with several other approaches in the literature. According to the results, the suggested method is capable of efficiently acquiring features at test time in a cost- and context-aware fashion.