Key source code for the article: "A Cost-Optimized 5-Protein Panel Revolutionizes Systemic Lupus Erythematosus Diagnosis"

This repository contains the core R analysis code for the manuscript "A Cost-Optimized 5-Protein Panel Revolutionizes Systemic Lupus Erythematosus Diagnosis".
Study summary:
Using plasma protein data from 544 SLE cases and 48,036 controls in the UK Biobank, we applied a balanced case-control sampling (BCCS) approach with LASSO regression to identify 35 high-confidence SLE-associated proteins. A protein risk score (ProtRS) derived from these proteins achieved excellent diagnostic performance (AUC = 0.91), outperforming polygenic risk scores and clinical factors alone. A cost-optimized 5-protein panel (TRIM21, SOD2, KLK3, IL15, ADIPOQ) retained high accuracy (AUC = 0.82) while reducing testing costs by ~87%. Population attributable fraction (PAF) analysis further underscored the dominant contribution of ProtRS to SLE burden.
Code contents:
The scripts (numbered 1–10) cover the core analysis pipeline:
• Data preparation, scaling, and mean imputation.
• 10,000 LASSO iterations with BCCS to select stable protein biomarkers.
• Computation of ProtRS and evaluation of the 35-protein model (1,000 BCCS).
• Comparison with polygenic risk score (PRS) and clinical risk factors.
• Integration of risk factors and categorical variable creation.
• Cost‑optimization by incremental protein addition.
• Population attributable fraction (PAF) calculations (single‑ and multi‑variable).
• Reproducibility check between 10,000 and 1,000 LASSO runs.
• Sensitivity analysis comparing MICE, median, and kNN imputation methods.
All scripts are fully reproducible and require the R packages listed in the README.
The UK Biobank data are not included due to access restrictions; please refer to the manuscript for data application procedures.