Abstract:

Infection classification is the key for choosing the proper treatment plans. Early determination of the causative agents is critical for disease control. Host responses analysis can detect variform and sensitive host inflammatory responses to ascertain the presence and type of the infection. However, traditional host-derived inflammatory indicators are insufficient for clinical infection classification. Fingerprints-based omic analysis has attracted increasing attention globally for analyzing the complex host systemic immune response. A single type of fingerprints is not applicable for infection classification (area under curve (AUC) of 0.550-0.617). Herein, an infection classification platform based on deep learning of dual plasma fingerprints (DPFs-DL) is developed. The DPFs with high reproducibility (coefficient of variation <15%) are obtained at low sample consumption (550 nL native plasma) using inorganic nanoparticle and organic matrix assisted laser desorption/ionization mass spectrometry. A classifier (DPFs-DL) for viral versus bacterial infection discrimination (AUC of 0.775) and coronavirus disease 2019 (COVID-2019) diagnosis (AUC of 0.917) is also built. Furthermore, a metabolic biomarker panel of two differentially regulated metabolites, which may serve as potential biomarkers for COVID-19 management (AUC of 0.677-0.883), is constructed. This study will contribute to the development of precision clinical care for infectious diseases.

Figure 1 Overall schematics for extraction of dual plasma fingerprints toward rapid infection classification by deep learning. a) Patient enrollment and plasma collection. Total number of study subjects (n = 318), including 187 viral-positive and 131 bacterial-positive. b) Data acquisition by LDI MS, including metabolic fingerprints and protein fingerprints. Only 1 µL of extracted plasma and 1 µL of diluted plasma sample were loaded on a microarray without any labeling for metabolic fingerprints and protein fingerprints for metabolic fingerprints, respectively. Then, matrix was directly loaded on the dried sample. Then, a Nd:YAG laser (355 nm) was irradiated onto the analyte microarray, facilitating NP enhanced and CHCA assisted LDI process in obtaining raw mass spectra. c) Deep learning of dual plasma fingerprints (DPFs-DL) for infection classification.

Figure 2 High-performance dual plasma fingerprints of infection by nanoparticle enhanced and CHCA assisted LDI MS. a) Scanning electron microscopy (SEM) image and Transmission electron microscopy (TEM) image (inset) of the nanoparticles (the scale bars were 100 nm). b,c) Intra-chip reproducibility study of b) standard metabolite mixture (valine, Val; lysine, Lys; glucose, Glc; and lactose, Lac) and c) protein mixture (bradykinin (1–7), Bra; angiotensin II, AngII; angiotensin I, AngI; and renin substrate, Res) by 5 independent replicates per sample. d) DPFs data matrix layout, including raw MS as input data, fingerprints of multi-center (uncorrected_MPFs and uncorrected_PPFs), batch effect correction, and data fusion (dual plasma fingerprints, DPFs). e) Typical mass spectra of metabolic and proteomic fingerprints for plasma samples of bacterial-positive and viral-positive patients. f) Batch-effect correction in MPFs and PPFs: plasma samples were collected from ten centers, and fingerprints were generated on one platform. g) Alignment score quantifying batch-effect correction. h) Heat map of DPFs for 301 samples (175 viral-positive and 126 bacterial-positive), using 146 m/z signals (123 for MPFs and 23 for PPFs) through data processing.

Figure 3 SARS-CoV-2 diagnosis and metabolic-biomarker discovery. a,b) Predicted score using DPFs-DL for SARS-CoV-2 diagnosis in a) train set (n = 77; 22/55, IFVs/SARS-CoV-2) and b) test set (n = 40; 28/12, IFVs/SARS-CoV-2) with cutoff of 0.6859. c) Metabolic-biomarker discovery. Two biomarkers were identified with accurate mass identification from six metabolites with FDR p-value <0.05 and paired adducts with Pearson correlation >0.6. d) Heatmap of the ion adduction peaks for two biomarkers. e,f) Scatter diagrams of paired adducts of biomarkers in IFVs and SARS-CoV-2 patients, including Leu and Glc. Levels of biomarkers were presented as normalized intensities. g) ROC curve of Leu with AUC of 0.883 and Glc with AUC of 0.677.

DOI: 10.1002/smll.202206349