Infectious diseases are significant causes of morbidity and mortality worldwide as well as spreading rapidly (Allerberger and Kern, 2020), with the complicated and multi-sites of occurrence. The annual incidences of meningitis, sepsis, and respiratory tract infections (RTIs) are approximately 850,000, 5.68 million, and 10 million in China, with a high mortality rate caused by confusing identification of the pathogen, complex etiology, and clinical decision challenges (Bodilsen et al., 2021; Wu et al., 2021). Therefore, rapid identification and appropriate diagnosis of pathogens in infectious disease management are of great clinical importance.

Currently, traditional detection of pathogenic microorganisms mainly utilizes morphology and molecular detection based on culture, which are basic tools and gold standard for clinical results confirming (Pliakos et al., 2018), especially the detection of bacteria and fungi. However, culture-based pathogen identification also has shortcomings, such as low detection accuracy (Parize et al., 2017), long culture cycle, high requirements on personnel operation, and inevitable deviations in the culture of unculturable microbes (Srivastava and Prasad, 2023). Thus, these methods cannot meet the current needs for rapid and accurate infection diagnosis; as a result, metagenomic next-generation sequencing (mNGS) has emerged as a clinical complement for gold-standard culture or even more clinically needed.

Rapid mNGS testing is a culture-independent, rapid detection of unknown infections for body fluids. It mainly includes short-read sequencing represented by Illumina platform and long-read sequencing represented by Nanopore platform (Mitchell and Simner, 2019; Zhang et al., 2022). Early studies have shown that mNGS can reliably detect and identify causative or emerging microorganisms in unexplained infectious disease syndromes (Li et al., 2019; Diao et al., 2022) and facilitate adjustment of empiric antibiotic therapy (Feng et al., 2023). The Illumina sequencing platform has been widely used clinically, but its short read length limits applicability to genomes. With long-read sequencing, nanopore-sequencing technology (NST) exhibits potential to compensate for this and meet the clinical needs for pathogen detection in high-viral-load clinical samples (Li et al., 2023) and species identification of Mycobacterium tuberculosis isolates (Smith et al., 2020). Therefore, NST may be a promising complement for sequencing platforms currently used in clinical settings. However, few studies have reported the diagnostic presentation of mNGS at different sites of infection, especially based on the nanopore platform.

Here, we designed a multi-center, observational study to investigate the detective and diagnostic performance of NST based on the analysis of 297 samples of suspected infections. With the inclusion of seven types of clinical fluid samples, we aimed to provide supporting evidence to evaluate the clinical impact of NST on real-world infection samples.

Fast and precise infection diagnosis makes a significant contribution to improving clinical management and patient prognosis, and mNGS is a promising tool to complement gold-standard culture in clinical practice. Although mNGS has been widely reported in the auxiliary diagnosis of infectious diseases, few studies have systematically evaluated the diagnostic performance of various fluid samples via NST in real-world infectious diseases, not to mention the comparison between sterile and microbial-colonized sites. This study included seven types of clinical fluids, from 10 clinical centers, and provided evidence to evaluate the effectiveness of nanopore sequencing in real-world clinical scenarios.

Nanopore sequencing exhibits fast detection (Gorzynski et al., 2022; Jain et al., 2022) and reduced turnaround time (Sheka et al., 2021) and enables both pathogen identification and AMR gene detection capabilities by unbiased rapidly sequencing. In addition, NST reduces the requirements of resource and technology, as well as capital and maintenance costs. Nanopore sequencing has been applied in genome assembly, full-length transcript detection, base modification detection, and other specialized areas, such as rapid clinical diagnoses and outbreak surveillance (Wang et al., 2021). The long-read length sequencing of NST could enhance the accuracy of real-time analysis of bioinformatics and, thus, provide opportunities for faster speed compared with other strategies using sequence-by-synthesis methods (Gu et al., 2019). In this regard, NST displays more potential for application as point-of-care testing (POCT) than Illumina platform (Ferreira et al., 2018); thus, the study on its true performance in clinical scenario could be of vital importance. Although mNGS is currently used for companion diagnosis, the diagnostic performances of NST in real-world clinical scenarios were rarely published. Most previous studies have only explored the diagnostic performance of mNGS with a single type or a limited number of samples, including bloodstream infections (sensitivity = 30.8%) (Hu et al., 2021), lower RTIs (sensitivity = 100%, specificity = 66%) (Yang et al., 2019) and meningitis (sensitivity = 41.7%, specificity = 78.6%) (Gao et al., 2021). In this study, the diagnostic performances in multiple types of samples were calculated based on a composite standard ( Table 3 and Supplementary Table S4 ). Even though the main disadvantages of mNGS, such as false positive and noise detection of background germs, were also observed in NST, efforts were made to discriminate positive detection with true pathogen by combined use of clinical adjudication in this study. This method has also been previously reported (Luo et al., 2023). The PPA and NPA of our results ( Table 3 ) exhibited comparable or even better performances with the previous publishing mentioned above. In addition, numbers of the latest Chinese expert consensus recommended mNGS as supplemental detection of emerging pathogens and acute infections. It supports that nanopore-based mNGS has potential to be used as a promising supplement for the existing detection in clinical diagnostic routines.

Considering that different types of samples characterize different sites of infection, we further analyzed the distributed features of body sites, including the microbial-colonized sites and originally sterile sites. A total of seven types of fluid samples were included in this study, three of them were collected from the respiratory and urinary tracts (BALF, sputum and urine), and the others were collected from sterile sites, where microbes could be detected only in pathological states. It could be observed that almost all types of samples exhibited a higher positive rate and clinical consistence with NST, especially abscess (PPA = 95.7%, NPA = 100%) and CSF (PPA = 88.2%, NPA = 95.7%). In addition, test accuracies of plasma samples (PPA = 77.6%, NPA = 66.7%) were also much better than previously reported (Hu et al., 2021) ( Table 3 ). However, statistical performances in SCE samples exhibited a volatile 95% CI, and the clinical consistency of NST was no better than that of culture method ( Supplementary Table S3 ), neither in positive nor negative tests. That possibly correlated with the low sample size and low pathogen load in SCE. For samples collected from microbial-colonized sites, NST exhibited 100% PPA but 14.3%–50.0% NPA. This may be due to the fact that, in real-world clinical scenarios, there is a very low probability that no microbes are detected (true negative samples) from the samples collected from microbial-colonized sites, in suspected infectious patients.

We further analyzed the distribution and clinical relevance of detected microorganisms in different body sites. There were more varieties of viral and fewer varieties of fungal pathogens detected from the aseptic body sites, compared with respiratory and urinary tracts ( Figures 4 , 5 ). As for clinically relevant bacteria, the top 3 frequently detected species were Klebsiella pneumoniae, E. coli and Staphylococcus aureus, from sterile body sites, while Mycobacterium tuberculosis, E. coli and Pseudomonas aeruginosa, from microbial-colonized sites. These species were in line with the results of several previous studies that investigated pathogens detected from plasma (Eichenberger et al., 2022), SCE (Liu et al., 2023), abscess (Zhao et al., 2021) and respiratory tract (Jin et al., 2022). Among the above species, there was a high concordance in detection of E. coli, P. aeruginosa and S. aureus between NST and culture. Therefore, for common clinical pathogens, it was enough to detect them through gold-standard culture, but it should be mentioned that NST has a faster diagnosis of pathogens than culture (Gu et al., 2021). As for the number of pathogen-derived reads, we recalculated the number of detected reads in each sample, and the top 10 identified pathogens were shown in Supplementary Figure S1B . Generally, the number of reads in bacterial pathogens could be considerably higher than that in viral pathogens. For bacterial pathogen which could be detected from both sterile and microbial-colonized body sites, the number of reads detected from the latter sites was greater. Interestingly, the number of detected reads in K. pneumoniae in samples collected from sterile body sites could reach 4693.6 ± 1895.0, which was greater than that from the microbial-colonized sites. The varied distribution of the numbers of pathogen-derived reads suggests the complexity of infectious diseases in the real-world clinical scenario, and it is of vital importance with clinical adjudication of NST detection based on the existing technical level.

On the other hand, there are also a number of pathogens better detected with NST than culture, such as Mycobacterium tuberculosis, human betaherpesvirus, Torque teno virus, and Pneumocystis jirovecii. It has been reported that NST can detect fastidious or unculturable microbes (Jin et al., 2022), identify complex infections (Duan et al., 2021), and assist diagnosis of conditional pathogens (Wang et al., 2012). Our study demonstrated that NST performed better than culture test in the detection of fastidious pathogens such as Mycobacterium spp., with additional detection of nineteen cases of MTB and six cases of NTM in culture-negative samples. In addition, common clinical conditioning pathogens, including Candida (Wang et al., 2012), Klebsiella pneumoniae(Chen et al., 2023a), Pneumocystis jiroveci (Wang et al., 2022), Epstein-Barr virus, and Adenovirus were also identified with NST in this study, which were further determined as pathogenic bacteria in combination with clinical symptoms, signs and hemograms of patients. Thus, NST can assist in the precisely and early diagnosis of fastidious and conditional pathogens and serve as an essential supplementary test for the gold standard test. Meanwhile, it should be noticed that NST-detected microbes could be pathogenic or colonized. Therefore, whether the detected microbes are clinically relevant still requires clinical identification and decision making by clinicians.

In addition to the identification of microorganisms, the predictive performance of NST in drug resistance was also notable in this study, which was only reported by a few previous publications (Serpa et al., 2022; Chen et al., 2023b; Xiao et al., 2023). In this study, among 56 microbial strains positive with AMR genes detected by NST only 33 of them were successfully cultured and detected by AST, as shown in Figure 5B . There were 58.9% (33 of 56) microbial strains failed to be cultured but were adjudicated as pathogens by clinicians. It should be noted that NST showed a higher detection rate of positive AMR genotype compared with the culture method, which could be actual drug resistance or positive genotype but negative phenotype. Also, it would be a pity that those hypotheses could not be testified because of the negative culture in this study. The failure of gold-standard culture would contribute to the unconscious of neither microbial identification nor drug resistance phenotype. From this perspective, AMR prediction by NST could provide an alternative strategy to infer antimicrobial susceptibility for clinicians and, thus, provide a reference for the selection of clinical treatment scheme, especially for patients with negative culture results. Further study could design prospective experiments for consistency comparisons between AMR genotype and drug resistance phenotype.

This study has the following limitations. First, due to the competitive enrollment of samples, the number of each type of clinical fluid was not consistent, ranging from 23 to 109. Second, the gold-standard culture was used for clinical routine tests of bacterial and fungal pathogens but with difficult application for clinical virus detection. Moreover, this study mainly took microorganisms inherited with DNA into coverage, so RNA virus was not detected or analyzed in this study. Third, the positive rate of culture was relatively low, and the AMR genotype prediction failed to be testified in this study. Further observational studies with large sample size are still necessary to provide more evidence for exploring the diagnostic performance of NST in the clinical setting.

In conclusion, this study includes seven types of infectious fluid, from 10 clinical centers, and provides evidence for application of NST in real-world clinical scenarios. For samples collected from microbial-colonized sites, NST exhibits more varieties of detected species compared with culture, which needs to be clinically adjudicated. For originally sterile sites, excellent diagnostic performances of NST could support evidence for pathogen identification and antimicrobial treatment. Moreover, NST is potentially valuable in the diagnosis and management of fastidious pathogens, complex infections, and opportunistic pathogen infections, and it is also capable of providing a new strategy to predict antimicrobial susceptibility for clinicians. It should be believed that NST is a promising supplement for gold-standard culture in real-world clinical scenarios.

The datasets of Nanopore sequencing for this study have been deposited into CNGB Sequence Archive (CNSA) of China National GeneBank DataBase (CNGBdb) with accession number CNP0004703.

The studies involving humans were approved by Ethics Committee of The First Affiliated Hospital of Soochow University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

MZ: Data curation, Formal analysis, Project administration, Supervision, Visualization, Writing – original draft. YYZ: Project administration, Supervision, Writing – original draft. LC: Data curation, Formal analysis, Visualization, Writing – original draft. XY: Investigation, Methodology, Writing – original draft. TX: Investigation, Methodology, Writing – original draft. MF: Investigation, Methodology, Writing – original draft. YH: Validation, Writing – original draft. YZ: Validation, Writing – original draft. BZ: Validation, Writing – original draft. JC: Validation, Writing – original draft. JL: Project administration, Supervision, Writing – review & editing. DS: Funding acquisition, Project administration, Supervision, Writing – review & editing. SL: Conceptualization, Data curation, Formal analysis, Visualization, Writing – review & editing. CZ: Conceptualization, Writing – review & editing. WZ: Conceptualization, Funding acquisition, Resources, Writing – review & editing.