Overcoming collaboration barriers in quantitative trait loci analysis.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Expression quantitative trait loci (eQTL) mapping is crucial for understanding gene regulation.
• Current methods often require centralizing sensitive genetic data, posing privacy risks.
• Federated learning offers a privacy-preserving alternative but faces computational challenges.

Purpose of the Study:

• To introduce privateQTL, a novel computational framework for federated eQTL analysis.
• To demonstrate the feasibility of conducting multi-institutional eQTL studies without data sharing.
• To address privacy concerns in large-scale genetic association studies.

Main Methods:

• Leveraging secure multiparty computation (MPC) for distributed data analysis.
• Implementing a federated learning approach for eQTL mapping.
• Developing algorithms to aggregate results securely across multiple institutions.

Main Results:

• Successfully performed federated eQTL mapping across simulated or real institutional datasets.
• Demonstrated that privateQTL maintains statistical power comparable to traditional methods.
• Confirmed the privacy-preserving nature of the MPC-based approach.

Conclusions:

• privateQTL enables privacy-preserving, federated eQTL analysis across institutions.
• This method facilitates collaborative genomic research while safeguarding sensitive data.
• Future genetic analyses can benefit from secure, distributed computation frameworks.