Privacy-preserving Polygenic Risk Scoring using Homomorphic Encryption

Islam, Shaedul; Demirci, Hüseyin; Lenzini, Gabriele

doi:10.3384/ecp187021

Cited by 2 publications

(2 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our approach entails the propagation of encrypted data across three parties (clients with sensitive genetic data, modelers with existing PRS models, and evaluators to interpret PRS findings), preserving model and genomic privacy when communicating PRS results back to the patient [21]. This protocol was inspired by the iDASH 2022 secure genome analysis competition [21], and our solution addresses task 2, “Secure Model Evaluation on Homomorphically Encrypted Genotype Data.” While other studies have proposed homomorphic encryption of PRS models and provided proof-of-concept on limited artificial datasets [22], we take advantage of the scalability of FHE to preserve the privacy of robust PRS models that contain a larger number of significant SNPs. In addition to robust assessment with synthetic datasets to show how our model scales with increased SNP numbers, we apply our FHE-based protocol to a 110k-SNP PRS model for schizophrenia [23], which shows accuracy in predicting schizophrenia risk in a cohort of over 1200 individuals with minimal decrease in performance compared to non-encrypted PRS.…”

Section: Introductionmentioning

confidence: 99%

Homomorphic Encryption: An Application to Polygenic Risk Scores

Knight,

Yolou,

et al. 2024

Preprint

View full text Add to dashboard Cite

Background: Polygenic risk scores (PRS) have emerged as a powerful tool in precision medicine, enabling personalized risk assessments for complex diseases. However, using sensitive genomic data in PRS calculations raises concerns about privacy and security. Fully Homomorphic Encryption (FHE) offers a promising solution by allowing computations on encrypted data, preserving the privacy of both genomic information and PRS models. Results: In this study, we present a novel application of FHE for secure and private PRS calculations using the CKKS protocol within the Lattigo library. Our approach involves a three-party system: clients (doctors with sensitive genetic data), modelers developing a PRS (academics or a company), and evaluators (a "local hospital" running the models while maintaining data confidentiality). We demonstrate the feasibility and accuracy of our protocol by applying it to synthetic datasets of various sizes and a robust 110k-SNP model for schizophrenia. The normal PRS calculation results are essentially identical to the encrypted calculation: between the two resultsR2is .999 & MSE is 2.27 times 10-6. Moreover, while the encrypted calculation is roughly 1000 times slower than conventional non-encrypted ones (when only considering the core PRS calculation), it is quite feasible on a single-CPU node - e.g. running on ≈ 1100 individuals with ≈ 110k SNPS took six minutes and ≈ 65G memory on a laptop computer. In addition, we investigate the impact of encryption parameters (modulus) on this computational time and accuracy in detail. Conclusion: By enabling secure PRS calculations on encrypted genomic data, our approach addresses the pressing need for privacy-preserving solutions in the era of precision medicine. The ability to perform accurate risk assessments while maintaining patient confidentiality paves the way for broader adoption of PRS and personalized medicine in healthcare, particularly with the advent of large-scale computing power.

show abstract

Section: Introductionmentioning

confidence: 99%

Homomorphic Encryption: An Application to Polygenic Risk Scores

Knight,

Yolou,

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…This protocol was inspired by the iDASH 2022 secure genome analysis competition [21], and our solution addresses task 2, "Secure Model Evaluation on Homomorphically Encrypted Genotype Data." While other studies have proposed homomorphic encryption of PRS models and provided proof-of-concept on limited artificial datasets [22], we take advantage of the scalability of FHE to preserve the privacy of robust PRS models that contain a larger number of significant SNPs. In addition to robust assessment with synthetic datasets to show how our model scales with increased SNP numbers, we apply our FHE-based protocol to a 110k-SNP PRS model for schizophrenia [23], which shows accuracy in predicting schizophrenia risk in a cohort of over 1200 individuals with minimal decrease in performance compared to non-encrypted PRS.…”

Section: Introductionmentioning

confidence: 99%

Homomorphic Encryption: An Application to Polygenic Risk Scores

knight,

Li,

Jensen

et al. 2024

Preprint

View full text Add to dashboard Cite

Background: Polygenic risk scores (PRS) have emerged as a powerful tool in precision medicine, enabling personalized risk assessments for complex diseases. However, using sensitive genomic data in PRS calculations raises concerns about privacy and security. Fully Homomorphic Encryption (FHE) offers a promising solution by allowing computations on encrypted data, preserving the privacy of both genomic information and PRS models. Methods: In this study, we present a novel application of FHE for secure and private PRS calculations using the CKKS protocol within the Lattigo library. Our approach involves a threeparty system: clients (doctors with sensitive genetic data), modelers developing a PRS (academics or a company), and evaluators (a ”local hospital” running the models while maintaining data confidentiality). We demonstrate the feasibility and accuracy of our protocol by applying it to synthetic datasets of various sizes and a robust 110k-SNP model for schizophrenia. Results: The normal PRS calculation results are essentially identical to the encrypted calculation: between the two results R2 is .999 & MSE is 2.27 × 10−6. Moreover, while the encrypted calculation is roughly 1000 times slower than conventional non-encrypted ones (when only considering the core PRS calculation), it is quite feasible on a single-CPU node - e.g. running on ∼1100 individuals with ∼110k SNPS took six minutes and ∼65G memory on a laptop computer. In addition, we investigate the impact of encryption parameters (modulus) on this computational time and accuracy in detail. Conclusion: Our approach enables secure PRS calculations on encrypted genomic data, addressing the pressing need for privacy-preserving solutions in the era of precision medicine. The ability to perform accurate risk assessments while maintaining patient confidentiality paves the way for broader adoption of PRS and personalized medicine in healthcare, particularly with the advent of large-scale computing power.

show abstract

Privacy-preserving Polygenic Risk Scoring using Homomorphic Encryption

Cited by 2 publications

References 12 publications

Homomorphic Encryption: An Application to Polygenic Risk Scores

Homomorphic Encryption: An Application to Polygenic Risk Scores

Homomorphic Encryption: An Application to Polygenic Risk Scores

Contact Info

Product

Resources

About