Privacy-preserving multicenter differential protein abundance analysis with FedProt.

Area of Science:

• Proteomics
• Bioinformatics
• Computational Biology

Background:

• Quantitative mass spectrometry allows large-scale protein quantification.
• Collaborative analysis of distributed patient data enhances statistical power but poses privacy risks.

Purpose of the Study:

• Introduce FedProt, a novel privacy-preserving tool for collaborative differential protein abundance analysis.
• Enable secure, multi-institutional proteomic data analysis using federated learning.

Main Methods:

• Developed FedProt utilizing federated learning and additive secret sharing.
• Created two novel multicenter datasets (E. coli and human serum) for evaluation.
• Compared FedProt's accuracy against existing methods like DEqMS.

Main Results:

• FedProt achieved accuracy comparable to centralized analysis (DEqMS) with negligible differences (≤ 4×10⁻¹²).
• Meta-analysis methods showed significant divergence in results compared to centralized analysis.
• Demonstrated the efficacy of FedProt on both microbial and human serum datasets.

Conclusions:

• FedProt offers a secure and accurate solution for collaborative proteomic analysis.
• The tool effectively addresses privacy concerns in multi-institutional data sharing.
• FedProt represents a significant advancement in privacy-preserving computational biology.