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.

Predicting Dog Phenotypes from Genotypes.

We analyze dog genotypes (i.e., positions of dog DNA sequences that often vary between different dogs) in order to predict the corresponding phenotypes (i.e., unique observed characteristics). More specifically, given chromosome data from a dog, we aim to predict the breed, height, and weight. We explore a variety of linear and non-linear classification and regression techniques to accomplish these three tasks. We also investigate the use of a neural network (both in linear and non-linear modes) for breed classification and compare the performance to traditional statistical methods. We show that linear methods generally outperform or match the performance of non-linear methods for breed classification. However, we show that the reverse is true for height and weight regression. Finally, we evaluate the results of all of these methods based on the number of input features used in the analysis. We conduct experiments using different fractions of the full genomic sequences, resulting in input sequences ranging from 20 SNPs to ∼200k SNPs. In doing so, we explore the impact of using a very limited number of SNPs for prediction. Our experiments demonstrate that these phenotypes in dogs can be predicted with as few as 0.5% of randomly selected SNPs (i.e., 992 SNPs) and that dog breeds can be classified with 50% balanced accuracy with as few as 0.02% SNPs (i.e., 40 SNPs).

Generative Moment Matching Networks for Genotype Simulation.

The generation of synthetic genomic sequences using neural networks has potential to ameliorate privacy and data sharing concerns and to mitigate potential bias within datasets due to under-representation of some population groups. However, there is not a consensus on which architectures, training procedures, and evaluation metrics should be used when simulating single nucleotide polymorphism (SNP) sequences with neural networks. In this paper, we explore the use of Generative Moment Matching Networks (GMMNs) for SNP simulation, we present some architectural and procedural changes to properly train the networks, and we introduce an evaluation scheme to qualitatively and quantitatively assess the quality of the simulated sequences.