Background

Hydroxychloroquine (HCQ) retinal toxicity risk is partially mediated by CYP2D6 metabolizer status, with poor metabolizers accumulating higher tissue levels (Petri et al., Arthritis Rheumatol 2020;72:1894-1901). However, pharmacogenomic data is among the most sensitive patient information, creating barriers to its integration into clinical decision support systems.

Hypothesis

We hypothesize that a fully homomorphic encryption (FHE) scheme applied to CYP2D6 diplotype classifications can enable real-time dose adjustment computations on encrypted genotype data, yielding identical dose recommendations to plaintext analysis while maintaining zero-knowledge of individual genetic variants.

Testable Predictions

1. Computational equivalence: FHE-encrypted CYP2D6 activity scores processed through a dose-adjustment algorithm will produce bit-identical outputs to plaintext computation in ≥99.99% of cases.
2. Latency threshold: End-to-end encrypted dose computation will complete in <500ms using TFHE or CKKS schemes on commodity hardware, making it viable for point-of-care use.
3. Privacy guarantee: Information-theoretic analysis will confirm that no CYP2D6 allele information is recoverable from the encrypted computation trace.

Proposed Validation

• Implement FHE-wrapped dose calculator using Concrete-ML or OpenFHE
• Benchmark against 1,000 simulated CYP2D6 profiles (UM, EM, IM, PM)
• Compare plaintext vs. encrypted outputs for concordance
• Measure latency on standard clinical workstation hardware

Significance

If confirmed, this approach would remove the primary barrier to pharmacogenomic-guided HCQ dosing in multi-site clinical networks, enabling safer prescribing without centralized genetic databases.

References

• Petri M et al. Arthritis Rheumatol 2020;72(11):1894-1901
• Melles RB, Jorge AM. JAMA Ophthalmol 2020;138(4):e200370
• Chillotti I et al. TFHE: Fast Fully Homomorphic Encryption over the Torus. J Cryptol 2020;33:34-91
• Marmor MF et al. Ophthalmology 2016;123:1386-94