Background/Purpose:

Biomarkers for use in developing treatments and diagnostics for Systemic Lupus Erythematosus (SLE) are a large unmet need. The wide differential in patient progression and outcome for SLE calls for the identification of informative biomarkers for use in patient stratification and determining course of treatment. In this study we used Berg Interrogative Biology®, a platform technology integrating multi-omic (metabolomics, lipidomics and proteomics) and artificial intelligence (bAIcis®) technologies to discover serum and urine based candidate markers of lupus.

Methods:

This study was conducted using retrospectively collected and clinically annotated serum and urine samples from 166 patients (90 African American and 71 Caucasian). Medical data included demographic data, ACR classification criteria, SLICC-damage index, SLEDAI disease activity scores, lab data, and medication information. Omics data for Mass Spectrometry included serum proteomics, metabolomics and lipidomics, and urine proteomics and metabolomics. As part of Berg Interrogative Biology® technology, all clinical and omics datasets were processed and fully integrated in a harmonized dataset. The multi-omic/clinic dataset was analyzed by Berg Artificial Intelligence technology (bAIcis®) that uses data-driven methods to identify panels of Lupus candidate biomarkers, each with a target area under the ROC curve (AUC) of 0.8 with the minimal combination of up to 6 biomarkers. Biomarker panels were analyzed separately for each biomatrix. bAIcis® provided (1) a summary table with individual AUC, panel AUC, panel power, and number of samples participated; (2) a panel ROC curve; and (3) a diagnostic table with statistics: sensitivity, specificity, positive predicted value (PPV), negative predicted value (NPV), and odds ratio.

Results:

Biomarker panels in serum and urine were discovered for lupus vs non-lupus (2 biomarkers in serum with AUC 0.836 and 5 in urine with AUC 0.805), renal disease vs no renal disease (2 biomarkers in serum with AUC 0.848 and 2 in urine with AUC 0.844), scleroderma vs. non-scleroderma (2 biomarkers in serum with AUC 0.826 and 2 in urine with AUC 0.705), scleroderma vs. lupus (5 biomarkers in serum with AUC 0.831 and 3 in urine with AUC 0.771), SLICC stage <2 vs >=2 (4 biomarkers in serum with AUC 0.829 and 2 in urine with AUC 0.77), SLEDAI score <6 vs >=6 (2 biomarkers in serum with AUC 0.809 and 2 in urine with AUC 0.641), ANA (1 biomarkers in serum with AUC 0.604 and 2 in urine with AUC 0.73), and drug efficacy for Mycophenolate (2 biomarkers in serum with AUC 0.847 and 1 in urine with AUC 0.933).

Conclusion:

Biomarker panels with AUC > 0.8 and power > 0.8 will be pursued in further prospective clinical study with a larger subject number. The urine and serum biomarker panels for lupus vs no lupus, renal disease vs no renal disease, scleroderma vs no scleroderma, scleroderma vs lupus, SLICC stage, and drug efficacy for Mycophenolate are fit for further validation. Integrating multi-omic analysis with artificial intelligence identified several biomarker panels that meet numerous unmet needs for the identification and clinical stratification of Systemic Lupus Erythematosus.

ACR Meeting Abstracts - https://acrabstracts.org/abstract/systemic-lupus-erythematosus-biomarkers-identified-using-multi-omic-and-artificial-intelligence-analysis-through-interrogative-biology/