Improving genetic risk prediction across diverse population by disentangling ancestry representations

Abstract

Risk prediction models using genetic data have seen increasing traction in genomics. However, most of the polygenic risk models were developed using data from participants with similar (mostly European) ancestry. This can lead to biases in the risk predictors resulting in poor generalization when applied to minority populations and admixed individuals such as African Americans. To address this bias, largely due to the prediction models being confounded by the underlying population structure, we propose a novel deep-learning framework that leverages data from diverse population and disentangles ancestry from the phenotype-relevant information in its representation. The ancestry disentangled representation can be used to build risk predictors that perform better across minority populations. We applied the proposed method to the analysis of Alzheimer’s disease genetics. Comparing with standard linear and nonlinear risk prediction methods, the proposed method substantially improves risk prediction in minority populations, including admixed individuals, without needing self-reported ancestry information.

Highly parameterized polygenic scores tend to overfit to population stratification via random effects

“…Given the demonstrated utility of PGSs in identifying high-risk individuals across diseases, our work suggests that part of this success may be owing to the PGS capturing population structure, which implies the usefulness of population structure as a predictive signal. One promising direction is the development of methods that accurately disentangle ancestry-relevant and phenotype-relevant signals, thereby leveraging the utility of both to improve PGS prediction power (40). Ultimately, the inclusion of diverse individuals in GWAS and in training PGSs will improve personalized genomic predictions.…”

FairPRS: adjusting for admixed populations in polygenic risk scores using invariant risk minimization

“… 9 Advances in machine learning such as using transfer learning-based methods 10 and deep learning based methods have been applied to make PRS more portable across ancestries. 11 However, either some of these methods assume part of the background genome is still of European origin 7 , 10 or consider pre-computed associated markers as input to reduce search space which can contain significant bias or spurious associations.…”