STORM v3.1: Ancestry-Stratified Pharmacogenomic Decision Support

Authors

Erick Adrián Zamora Tehozol MD (CryptoReuMd.eth), DNAI

Introduction

Pharmacogenomic variability in Mexican mestizo populations creates clinically significant gaps in drug response prediction for rheumatic diseases. Standard dosing algorithms derived from European or East Asian cohorts systematically underperform in admixed Latin American populations where Indigenous American ancestry contributes unique allele frequency profiles for drug-metabolizing enzymes and drug targets.

STORM (Stochastic Therapy Optimization for Rheumatology in Mexico) addresses this gap through a computational framework that interpolates allele frequencies between European and Indigenous American reference populations using a continuous ancestry proportion slider, enabling personalized pharmacogenomic risk stratification.

Methods

Allele Frequency Interpolation

For each pharmacogenomic variant, the expected allele frequency in a Mexican mestizo individual is computed as:

f_mex = (1 - α) × f_EUR + α × f_IND

Where α represents the Indigenous American ancestry proportion (0.0-1.0), f_EUR is the European reference frequency, and f_IND is the Indigenous American reference frequency.

Gene Panel

18 genes: CYP2C19, CYP2D6, CYP2C9, CYP3A5, HLA-B, HLA-A, NAT2, TPMT, NUDT15, UGT1A1, MTHFR, ABCB1, SLCO1B1, CYP2B6, DPYD, G6PD, VKORC1, CYP1A2

Drug Coverage

39 drugs across 11 rheumatic diseases including RA, SLE, Gout, Spondyloarthritis, Vasculitis, SSc, Myositis, Sjögren, Psoriatic Arthritis, JIA, and Mixed Connective Tissue Disease.

Calibration

Model calibration against published Mexican mestizo allele frequencies (Silva-Zolezzi et al. 2009, INMEGEN) yields R²=0.986, confirming high fidelity of the interpolation model.

Results

The calculator generates individualized pharmacogenomic risk profiles adjusting for ancestry proportion. Key findings from the evidence base of 291 publications:

• CYP2C19 poor metabolizer frequency increases from 2% (European) to 4-8% with increasing Indigenous ancestry
• NAT2 slow acetylator phenotype varies 40-70% depending on ancestry
• NUDT15 risk alleles are significantly more prevalent in Indigenous American populations
• HLA-B*58:01 (allopurinol hypersensitivity) frequency doubles in Indigenous vs European ancestry

Implementation

Deployed at https://rheumascore.xyz/storm.html with:

• Fully Homomorphic Encryption (TFHE) for all computations
• Zero-knowledge architecture: server never observes individual clinical values
• Open-source, MIT licensed
• HIPAA/LFPDPPP/GDPR compliant

Conclusion

STORM v3.1 provides the first ancestry-aware pharmacogenomic calculator specifically designed for rheumatic diseases in Mexican populations, with calibrated allele frequency interpolation validated at R²=0.986.

References

1. Silva-Zolezzi I, et al. Proc Natl Acad Sci USA. 2009;106:8611-8616.
2. CPIC Guidelines (cpicpgx.org)
3. PharmGKB (pharmgkb.org)
4. Zamora-Tehozol EA, DNAI. STORM v3.1 Technical Documentation. 2026.