A noninvasive, easy operation, and accurate diagnostic protocol is highly demanded to assess systemic lupus erythematosus (SLE) activity during pregnancy, promising real-time activity monitoring during the whole gestational period to reduce adverse pregnancy outcomes. Here, machine learning of serum metabolic fingerprints (SMFs) is developed to assess the SLE activity for pregnant women. The SMFs are directly extracted through a hollow-cobalt oxide/carbon (Co3 O4 /C)-composite-assisted laser desorption/ionization mass spectrometer (LDI MS) platform. The Co3 O4 /C composite owns enhanced light absorption, size-selective trapping, and better charge-hole separation, enabling improved ionization efficiency and selectivity for LDI MS detection toward small molecules. Metabolic fingerprints are collected from ≈0.1 µL serum within 1 s without enrichment and encoded by the optimized elastic net algorithm. The averaged area under the curve (AUC) value in the differentiation of active SLE from inactive SLE and healthy controls reaches 0.985 and 0.990, respectively. Further, a simplified panel based on four identified metabolites is built to distinguish SLE flares in pregnant women with the highest AUC value of 0.875 for the blind test. This work sets an accurate and practical protocol for SLE activity assessment during pregnancy, promoting precision diagnosis of disease status transitions in clinics.
Keywords: cobalt oxide; diagnostics; mass spectrometry; metabolites; systemic lupus erythematosus.
© 2022 Wiley-VCH GmbH.