PRANA: A Deep Learning Method for Adapting Polygenic Risk Scores to Diverse Ethnic Groups
Abstract
Polygenic risk scores (PRSs), which quantify inherited susceptibility to complex traits and diseases, have emerged as valuable tools for risk stratification and precision medicine. Despite their promise, PRS developed on European cohorts often demonstrate substantially reduced predictive accuracy in non-European populations, due to differences in genetic architecture. The disproportionate representation of European ancestry cohorts in genome-wide association studies (GWAS) leads to inequitable deployment of PRS technologies across diverse populations. Here, we introduce PRANA (Polygenic Risk Adaptation via Neural-network Architecture), a deep learning framework that adapts an existing PRS developed on one population to other ancestries. Unlike methods that require large-scale GWAS in the target population, PRANA leverages pre-trained PRS models derived from European cohorts and adapts them using modestly sized cohorts from the target population. We evaluated PRANA on seven complex traits in South Asian, East Asian and Ashkenazi Jewish populations, as well as in selected smaller East Asian subpopulations where the scarcity of training data poses a particular challenge. PRANA mostly improved predictive performance of the baseline PRS models by 5%-20% in terms of effect size (β) and Nagelkerke’s R², and, in most cases, outperformed existing cross-ancestry multi-PRS approaches. These results highlight PRANA as a scalable and practical strategy to reduce disparities in genomic risk prediction and advance the equitable application of PRS in diverse populations.
Related
- From Monogenic Variants to Polygenic Risk Scores in Inherited Cardiomyopathies: Clinical Utility, Ancestry Gaps, and Future Implementation in Arab Populations
- From Nagelkerke's R2 to Liability-Scale Variance Explained for Polygenic Scores
- BLocPRS: locus-aware ensembles of polygenic scores provide accurate linear prediction
- GeneSIS: enhancing transferability of polygenic scores with variant-level gene-by-sex interaction effects
- Education-related polygenic scores show different associations with youth mental health in parent-offspring trios
BibTeX
@misc{levi2026prana,
title = {{PRANA: A Deep Learning Method for Adapting Polygenic Risk Scores to Diverse Ethnic Groups}},
author = {Hagai Levi and Qin Wang and Manjeet K. Bolla and Joe Dennis and Irene L. Andrulis and Natalia Antonenkova and Chun Hang Au and Annelie Augustinsson and Laura E. Beane Freeman and Sabine Behrens and Marina Bermisheva and Clara Bodelon and Natalia V. Bogdanova and Stig E. Bojesen and Hermann Brenner and Ian W. Brock and Thomas Brüning and Helen Byers and Nicola J. Camp and Jose E. Castelao and Ji-Yeob Choi and Wendy K. Chung and Nbcs Collaborators and Sarah V. Colonna and Fergus J. Couch and Kamila Czene and Mary B. Daly and Peter Devilee and Thilo Dörk and A. Heather Eliassen and Mikael Eriksson and D. Gareth Evans and Peter A. Fasching and Kierstin Faw and Manuela Gago-Dominguez and Montserrat García-Closas and Christopher A. Haiman and Ute Hamann and Mikael Hartman and Vikki Ho and Peh Joo Ho and Maartje J. Hooning and Reiner Hoppe and Sacha J. Howell and Abctb Investigators and Sgbcc Investigators and Hidemi Ito and Motoki Iwasaki and Anna Jakubowska and Helena Jernström and Vijai Joseph and Rudolf Kaaks and Daehee Kang and Elza K. Khusnutdinova and Sanja Kiprijanovska and Stella Koutros and Vessela N. Kristensen and Allison W. Kurian and Ava Kwong and Diether Lambrechts},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.07.12.26357860},
url = {https://hal.science/hal-05748767v1},
note = {CC BY-ND},
}
RIS
TY - UNPB
TI - PRANA: A Deep Learning Method for Adapting Polygenic Risk Scores to Diverse Ethnic Groups
AU - Hagai Levi
AU - Qin Wang
AU - Manjeet K. Bolla
AU - Joe Dennis
AU - Irene L. Andrulis
AU - Natalia Antonenkova
AU - Chun Hang Au
AU - Annelie Augustinsson
AU - Laura E. Beane Freeman
AU - Sabine Behrens
AU - Marina Bermisheva
AU - Clara Bodelon
AU - Natalia V. Bogdanova
AU - Stig E. Bojesen
AU - Hermann Brenner
AU - Ian W. Brock
AU - Thomas Brüning
AU - Helen Byers
AU - Nicola J. Camp
AU - Jose E. Castelao
AU - Ji-Yeob Choi
AU - Wendy K. Chung
AU - Nbcs Collaborators
AU - Sarah V. Colonna
AU - Fergus J. Couch
AU - Kamila Czene
AU - Mary B. Daly
AU - Peter Devilee
AU - Thilo Dörk
AU - A. Heather Eliassen
AU - Mikael Eriksson
AU - D. Gareth Evans
AU - Peter A. Fasching
AU - Kierstin Faw
AU - Manuela Gago-Dominguez
AU - Montserrat García-Closas
AU - Christopher A. Haiman
AU - Ute Hamann
AU - Mikael Hartman
AU - Vikki Ho
AU - Peh Joo Ho
AU - Maartje J. Hooning
AU - Reiner Hoppe
AU - Sacha J. Howell
AU - Abctb Investigators
AU - Sgbcc Investigators
AU - Hidemi Ito
AU - Motoki Iwasaki
AU - Anna Jakubowska
AU - Helena Jernström
AU - Vijai Joseph
AU - Rudolf Kaaks
AU - Daehee Kang
AU - Elza K. Khusnutdinova
AU - Sanja Kiprijanovska
AU - Stella Koutros
AU - Vessela N. Kristensen
AU - Allison W. Kurian
AU - Ava Kwong
AU - Diether Lambrechts
PY - 2026
DA - 2026/09/14
PB - medRxiv
DO - 10.64898/2026.07.12.26357860
UR - https://hal.science/hal-05748767v1
AB - Polygenic risk scores (PRSs), which quantify inherited susceptibility to complex traits and diseases, have emerged as valuable tools for risk stratification and precision medicine. Despite their promise, PRS developed on European cohorts often demonstrate substantially reduced predictive accuracy in non-European populations, due to differences in genetic architecture. The disproportionate representation of European ancestry cohorts in genome-wide association studies (GWAS) leads to inequitable deployment of PRS technologies across diverse populations. Here, we introduce PRANA (Polygenic Risk Adaptation via Neural-network Architecture), a deep learning framework that adapts an existing PRS developed on one population to other ancestries. Unlike methods that require large-scale GWAS in the target population, PRANA leverages pre-trained PRS models derived from European cohorts and adapts them using modestly sized cohorts from the target population. We evaluated PRANA on seven complex traits in South Asian, East Asian and Ashkenazi Jewish populations, as well as in selected smaller East Asian subpopulations where the scarcity of training data poses a particular challenge. PRANA mostly improved predictive performance of the baseline PRS models by 5%-20% in terms of effect size (β) and Nagelkerke’s R², and, in most cases, outperformed existing cross-ancestry multi-PRS approaches. These results highlight PRANA as a scalable and practical strategy to reduce disparities in genomic risk prediction and advance the equitable application of PRS in diverse populations.
ER -