Abstract

Infective endocarditis (IE) continues to pose significant clinical challenges as a life-threatening condition associated with 30% mortality. The current diagnostic criteria, the 2023 Duke-International Society for Cardiovascular Infectious Diseases (ISCVID) criteria, present diagnostic challenges due to complex processes. Blood culture remains a cornerstone of IE diagnosis, enabling identification of the causative microorganism and guiding targeted antibiotic therapy. However, results typically take 2–5 days, significantly delaying critical treatment decisions. To overcome these limitations, we developed a nanoparticle-enhanced laser desorption/ionization mass spectrometry (NPELDI MS) platform capable of acquiring serum metabolic fingerprints (SMFs). When integrated with machine learning algorithms, this platform achieves accurate IE diagnosis (area under the curve (AUC) = 0.882) and rapid streptococcal classification within 10 min. Notably, our platform enables simultaneous IE diagnosis and classification via a single assay free of culture process. This integrated approach addresses the critical unmet need in IE management, offering transformative potential for timely therapeutic decision-making and improved patient outcomes.

Scheme 1. Overview of the study design. Serum metabolic fingerprints (SMFs) were obtained by nanoparticle-enhanced laser desorption/ionization mass spectrometry (NPELDI MS). Biomarker identification and model construction were performed based on machine learning to enable the diagnosis and classification of infective endocarditis (IE).

Fig. 1. Comprehensive metabolic landscape of IE. (A) Clinical parameters of the study cohort, including sex and age (years) of healthy controls (HC, blue) and IE patients (pink). (B) Digital image of the microarray showing printed matrix regions (orange). Scale bar: 5 mm. (C) Representative mass spectra at the m/z range of 100–500 from serum samples of HC and IE patients. (D) Intra-group similarity score frequency distributions based on SMFs within the same group. (E) Heatmap of 146 m/z features across 72 serum samples after preprocessing (logarithmically scaled). (F) Volcano plot identifying 23 upregulated and 13 downregulated metabolites in IE patients (p < 0.05, fold change >1.25 or < 0.80). (G) Principal component analysis (PCA) demonstrating distinct clustering of IE and HC groups. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

https://doi.org/10.1016/j.bios.2026.118425