Background Lupus nephritis (LN) is a leading cause of mortality in patients with systemic lupus erythematosus (SLE), and the accurate classification of renal pathological subtypes is crucial for reducing mortality rates and improving long-term prognosis. Renal biopsy is the gold standard for LN diagnosis and classification; however, it is invasive, costly, and difficult to use for repeated monitoring or in all patient populations. Methods This study established a non-invasive liquid biopsy platform based on surface-enhanced Raman spectroscopy (SERS), combined with supervised machine learning (random forest algorithm, leave-one-out cross-validation), using urine samples to achieve the diagnosis and pathological subtype classification of LN. Silver nanoparticles were used as SERS-active substrates to identify urinary biomarkers associated with LN. The study included both LN patients and those with nephrotic syndrome (NS). Machine learning algorithms were used to extract spectral features and build classification models to distinguish LN from NS. Additionally, SERS of different LN pathological subtypes were analyzed to clarify subtype-specific urinary molecular characteristics. Results The results showed that SERS combined with machine learning can reliably and noninvasively distinguish LN from NS, achieving an LN diagnostic accuracy of 93.55%, and can stratify the main pathological subtypes of LN. Conclusion This liquid biopsy strategy holds significant potential for non-invasive diagnosis, subtype classification, and personalized treatment decisions in LN.
Xia et al. (Tue,) studied this question.