Privacy-preserving biological age prediction over federated human methylation data using fully homomorphic encryption

Area of Science:

• Bioinformatics
• Computational Biology
• Genetics

Background:

• DNA methylation data are vital for estimating chronological and biological age.
• The epigenetic pacemaker (EPM) model infers biological age, requiring sensitive data protection.
• Existing privacy-preserving methods for EPM computation have limitations in communication complexity and scalability.

Purpose of the Study:

• To develop a new privacy-preserving protocol for epigenetic pacemaker (EPM) computation.
• To improve upon the privacy and computational efficiency of existing methods.
• To ensure data security in biological age estimation, particularly for medical applications.

Main Methods:

• Development of a novel privacy-preserving protocol for EPM computation.
• Utilization of a single server for secure computation, enhancing resilience against server corruption.
• Application of techniques from symbolic algebra and number theory to optimize the protocol.
• Implementation and performance evaluation of the new protocol.

Main Results:

• The new protocol offers analytical improvements in both privacy and computational complexity.
• Elimination of communication during secure computation, leading to significant runtime improvements.
• Demonstrated ability to produce results comparable to the standard EPM model.
• Substantial performance gains compared to previous privacy-preserving approaches.

Conclusions:

• The developed protocol enhances data security and efficiency for EPM-based biological age estimation.
• This approach is practical for large datasets and medical applications where patient privacy is critical.
• The protocol's generality suggests potential applications in other fields using expectation-maximization techniques.