Unveiling the Future of Infective Endocarditis Diagnosis: The Transformative Role of Metagenomic Next-Generation Sequencing in Culture-Negative Cases

Abstract

Culture-negative infective endocarditis (CNE) remains a significant diagnostic challenge in cardiology and infectious disease, often leading to delayed or empirical treatment. Metagenomic next-generation sequencing (mNGS) has emerged as a complementary diagnostic tool capable of identifying fastidious, unexpected, or novel pathogens without prior assumptions. This narrative review synthesizes evidence from 152 studies (2015–2024), evaluating mNGS within existing diagnostic frameworks for culture-negative IE. Compared to conventional diagnostics (blood cultures, PCR, 16 S rRNA sequencing), mNGS demonstrates enhanced detection capabilities for polymicrobial infections and rare pathogens, though methodological heterogeneity across studies precludes definitive performance comparisons. Performance varies substantially based on sample type, sequencing platform, and bioinformatic pipelines. Real-world applications reveal persistent challenges, including cost barriers, interpretive complexities in low-biomass samples, and contamination risks. Integration with host-response biomarkers and AI-driven interpretation platforms shows promise for advancing clinical utility. For mNGS to be effectively integrated into routine CNE care, standardization, regulatory clarity, and equitable implementation will be essential.

Introduction

Infective endocarditis (IE) is a severe infection of the heart endocardial layer, predominantly affecting heart valves. Firstly, explained by Sir William Osler in 1885, IE has historically posed significant diagnostic and therapeutic challenges. Despite advancements in medical science, the prognosis for IE remains grim, with in-hospital or three-month mortality rates ranging from 18 to 25%, a figure that has remained largely unchanged over the years [1, 2]. The burden of IE on healthcare systems is substantial, given its associated high morbidity and long-term complications, which significantly impair patients’ quality of life. Over recent decades, changes in the epidemiology of IE have been observed. The increase in elderly populations, along with the increasing prevalences of degenerative valvular diseases and prosthetic heart valve usage, has shifted the context of this condition. Staphylococci, particularly Staphylococcus aureus, have surpassed Streptococci as the predominant causative pathogens. Furthermore, emerging and re-emerging pathogens, including rare and fastidious organisms, have complicated the clinical picture [3]. For diagnostic challenges, culture-negative infective endocarditis (CNE) presents a unique set of challenges. CNE occurs when standard blood cultures fail to identify the causative organism, which happens in up to 30% of IE cases. The causes of CNE are multifactorial, including prior antibiotic therapy, which could suppresses bacterial growth in cultures; the presence of fastidious organisms that require specific growth conditions not met by routine culture techniques; and intracellular pathogens that evade detection [4]. The clinical implications of CNE are profound. Diagnostic delays are common, as traditional methods often yield inconclusive results, leading to prolonged empirical therapy with broad-spectrum antibiotics. This approach can contribute to antibiotic resistance and adverse drug reactions, further complicating patient management. Inappropriate or delayed treatment significantly impacts patient outcomes, underlining the necessity for more effective diagnostic strategies [1]. Comparatively, molecular diagnostic techniques have revolutionized the approach to diagnosing infectious diseases, including IE. These methods offer several advantages over traditional culture-based techniques, particularly in their ability to detect a wide range of pathogens with high sensitivity and specificity (Table 1). One of these methods, polymerase chain reaction (PCR) and 16 S rRNA sequencing have provided substantial improvements but are limited by the requirement for prior knowledge of the target pathogen and the scope of detectable organisms [5]. Metagenomic next-generation sequencing (mNGS) represents a significant advancement in molecular diagnostics. This technique enables for the extensive analysis of genetic material from clinical samples, identifying all present pathogens without prior knowledge. mNGS works by sequencing entire sample on nucleic acids, followed by bioinformatics analysis to identify the genetic sequences of potential pathogens [6, 7]. Unlike PCR and targeted sequencing, mNGS offers a broader detection spectrum, including viruses, fungi, bacteria, and parasites. Our current work aims to critically evaluate the mNGS application in the early and accurate diagnoses of CNE. This work is not a formal meta-analysis, but it synthesizes recent advancements and the best available resources, offering new insights into the diagnostic potential of mNGS. We compare mNGS with traditional diagnostic methods, discuss its clinical utility, and examine its impact on patient outcomes. Additionally, we identify current trends, research gaps, and propose future directions to optimize the mNGS utilization in clinical practice. The review provides a extensive exploration on the evolution of diagnostic techniques for IE (Table 2), with a focus on mNGS. It could serve as important resources for healthcare professionals, academics, and administrators about the potential of mNGS in improving the IE diagnoses and management. Through this synthesis, we hope to highlight the critical advancements and future possibilities that mNGS brings to the aspect of infectious diseases diagnostics.

Table 1.

Comparative sensitivity & specificity on mNGS vs. traditional diagnostic Methods

Diagnostic Method	Sensitivity	Specificity	Detection Time	Pathogen Coverage	Reference
Traditional Blood Culture	40–60%	High	Days to Weeks	Bacteria, Limited	[102]
PCR	70–80%	High	Hours to Days	Targeted Pathogens	[102, 103]
mNGS	90–95%	High	24–48 Hours	Comprehensive	[104]

Table 2.

Evolution of diagnostic techniques for infective endocarditis

Diagnostic Technique	Introduction Period	Key Advantages	Limitations	Examples of Pathogens Detected	References
Blood Cultures	Early 1900s	Gold standard; can identify a wide range of bacteria in IE; easy to implement.	Fails in 30% of cases, especially post-antibiotics; slow to yield results (days).	Staphylococcus aureus, Streptococcus viridans	[19]
Serology	1950s-1960s	Non-invasive; useful for fastidious organisms like Coxiella burnetii.	Requires specific tests for different organisms; lacks broad applicability.	Coxiella burnetii (Q fever)	[20]
Histopathology	1900s (Routine use began later)	Direct visualization of valvular tissue; provides definitive diagnosis.	Invasive; requires heart valve biopsy; not always feasible.	Any organism causing valve damage (e.g., bacterial, fungal)	[21, 22]
Polymerase Chain Reaction (PCR)	1980s-1990s	High sensitivity; detects DNA directly from clinical samples.	Relatively costly and limited to known pathogens with available primers.	Bartonella, Tropheryma whipplei	[23]
16S rRNA Sequencing	Late 1990s	Broadens bacterial identification range; effective for rare and slow-growing bacteria.	Limited to bacterial pathogens; cannot detect viruses or fungi.	Tropheryma whipplei, Bartonella species	[24]
Metagenomic Next-Generation Sequencing (mNGS)	2010s-Present	Rapid, comprehensive detection of bacteria, fungi, and viruses.	High cost, complex data interpretation, lack of standardized clinical protocols.	Broad range: bacterial, viral, fungal (including rare and fastidious organisms)	[25]

Methodology

A comprehensive literature search was conducted in accordance with PRISMA guidelines across PubMed, Scopus, and Web of Science to assess the role of metagenomic next-generation sequencing (mNGS) in diagnosing culture-negative endocarditis (CNE). The search spanned January 2015 to February 2024 using keywords such as ‘mNGS,’ ‘infective endocarditis,’ and ‘culture-negative endocarditis.’ Only human, English-language, peer-reviewed studies were included. Additional records were identified through manual screening of reference lists from guidelines (ESC, ACC/AHA), reviews, and meta-analyses. Eligible studies included original research and case series evaluating mNGS for CNE diagnosis, particularly those comparing it to conventional diagnostics (e.g., blood cultures, PCR, 16 S rRNA) or reporting on clinical outcomes, diagnostic performance, or cost-effectiveness. Exclusion criteria included non-human studies, non-English publications, and isolated case reports without outcome or comparative data. From 1,247 initial records, two independent reviewers screened titles/abstracts and performed full-text reviews, yielding 152 studies meeting inclusion criteria. Data were extracted using a standardized template capturing study design, patient population, diagnostic methods, performance metrics, clinical impact, and cost data. Subgroup analyses were performed based on sample type, sequencing platform, and bioinformatics approach. Study quality and risk of bias were assessed using the QUADAS-2 tool. Data synthesis included narrative review and pooled diagnostic estimates from high-quality studies, emphasizing comparisons between mNGS and conventional methods on their clinical utility. Heterogeneity was critically appraised with attention to methodological and cohort-level differences.

Evolution of Diagnostic Techniques for Infective Endocarditis

Early Diagnostic Methods

Blood cultures have long been considered the ideal method in diagnosing infective endocarditis (IE), Tables 2 and 3. This technique, while foundational, suffers from significant limitations. For example, studies indicate that blood cultures fail to identify pathogens in up to 30% of IE cases, particularly in those who receive prior antibiotics [8, 9]. This high rate of culture-negative results underscores the inadequacy of blood cultures in a subset of patients. Furthermore, blood cultures can take many days to yield results, delaying crucial treatment necessitating advanced diagnostic strategies. Comparatively, modern molecular techniques promise faster with highly precise diagnoses, emphasizing the needs for advancement beyond traditional methods (Fig. 1). For the role of serology and histopathology in IE diagnosis, serological tests and histopathological examinations have supplemented blood cultures in diagnosing IE. Serology is highly significant in detecting antibodies against fastidious microbes, like Coxiella burnetii, responsible for Q fever endocarditis [10–14]. Histopathology provides direct visual confirmation of infection through examination of excised heart valves. While these methods offer critical diagnostic insights, their utility is restricted by the availability of specific tests and the invasive nature of tissue sampling. The need for less invasive yet highly accurate diagnostic tools remain pressing. Historical case studies intensely demonstrate the difficulties posed by early diagnostic methods. For instance, before the advent of advanced diagnostics, many cases of IE were only definitively diagnosed post-mortem, revealing the diagnostic limitations of the time [15, 16]. These cases often involved extensive valvular damage by the time of diagnosis, emphasizing the critical importance of earlier detection methods. The contrast with current capabilities underscores the significant progress made and highlights ongoing gaps that require further innovation. However, continuous research effort led to the presentation of polymerase chain reaction (PCR) and 16 S rRNA sequencing represented a leap forward in molecular diagnostics for IE summarized in Table 2. PCR permits the specific sequences amplification, thus detecting pathogens directly from clinical sample with higher sensitivity. 16 S rRNA sequencing, which targets a conserved region of bacterial ribosomal RNA, broadens the detection scope to include a variety of bacteria [5]. Studies demonstrated that these methods significantly improve diagnostic accuracy in cases where traditional cultures fail, such as identifying Tropheryma whipplei and Bartonella species [17, 18].

Table 3.

Summary of current IE diagnosis and treatment strategies by major global cardiovascular organizations

Organization	Diagnostic Criteria	Diagnostic Methods	Treatment Strategies	Key Guidelines & Updates	Reference
American Heart Association (AHA)	Modified Duke Criteria	Blood cultures, Echocardiography (TTE/TEE), Serology	Empirical broad-spectrum antibiotics, Valve surgery in severe cases	2020 Update: Emphasis on early surgery for certain high-risk patients	[1]
European Society of Cardiology (ESC)	Modified Duke Criteria with emphasis on imaging	Blood cultures, Echocardiography, 18F-FDG PET/CT for prosthetic valve IE	Targeted antibiotic therapy based on culture results, Early surgery for complications	2015 Guidelines: Inclusion of 18F-FDG PET/CT in diagnostic criteria	[4]
British Society for Antimicrobial Chemotherapy (BSAC)	Modified Duke Criteria	Blood cultures, Echocardiography, Molecular methods (PCR)	Combination antibiotic therapy, Consideration for surgery	2020 Guidelines: Updated antibiotic regimens, More emphasis on molecular diagnostics	[70]
World Health Organization (WHO)	Adaptation of Duke Criteria	Blood cultures, Echocardiography, Emerging molecular diagnostics (mNGS)	Standardized antibiotic protocols, Surgical intervention in severe cases	2018 REASSURED: Promotion of molecular diagnostics in resource-limited settings	[71]
Infectious Diseases Society of America (IDSA)	Modified Duke Criteria	Blood cultures, Echocardiography, Histopathology	Empirical antibiotics, Custom therapy post-identification, Valve replacement	2015 Guidelines: Integration of newer diagnostic techniques, Aggressive surgical approach in high-risk patients	[1]

Fig. 1.

Principles and clinical application of mNGS in culture negative IE. Figure 1 illustrates the mNGS workflow for diagnosing CNE, from tissue sample processing and sequencing to bioinformatic analysis. Highlighting pathogen identification, phylogenetic analysis, and the clinical application of results for improving diagnosis, guiding therapy, and enhancing patient outcomes

mNGS in CNE Diagnosis

The persistent challenge of CNE, accounting for approximately 30% of IE cases [26], representing a critical diagnostic frontier where conventional methods reach their limitations. Prior antibiotic exposure, fastidious microorganisms, and intracellular pathogens collectively create a perfect diagnostic storm, leaving clinicians without actionable microbiological guidance. While PCR and 16 S rRNA sequencing offered incremental improvements, their targeted nature inherently restricts detection to predefined pathogens [27, 28]. Against this backdrop, mNGS emerges as a complement to traditional approaches, enabling comprehensive, unbiased pathogen detection directly from clinical specimens without prior cultivation [29]. This approach could enhance diagnostic possibilities for CNE by simultaneously interrogating bacterial, viral, fungal, and parasitic genomes within a single assay.

Technical Principles and Diagnostic Advantages

The core innovation of mNGS lies in its capacity to sequence all nucleic acids from host and microbial extracts in clinical samples. This hypothesis-free approach is particularly powerful for IE diagnostics, where causative agents range from common bacteria to exotic fastidious pathogens [30]. The analytical process integrates two critical phases: (i) Wet laboratory processing begins with optimal specimen selection. For IE, excised valvular tissue offers the highest diagnostic yield, as it harbors embedded pathogens protected from immune clearance. Blood samples, though less sensitive, provide a non-invasive alternative. Nucleic acid extraction must preserve both DNA and RNA to capture the full pathogen spectrum. Subsequent library preparation fragments nucleic acids and adds platform-specific adapters. A critical decision point involves selecting DNA-only sequencing (for bacteria, fungi, DNA viruses, parasites), RNA-only (for RNA viruses), or co-testing when the etiological category remains unknown, a flexibility indispensable for IE diverse microbiology [31, 32]. (ii) Bioinformatic analysis employs sophisticated computational pipelines to manage massive sequence datasets. After quality control, human-derived sequences (constituting > 99% of data in blood/tissue samples) are subtracted using reference genomes. The remaining sequences undergo alignment against comprehensive microbial databases, enabling pathogen identification at species or even strain levels. Advanced algorithms further characterize antimicrobial resistance genes and virulence factors, providing clinically actionable insights beyond mere detection [33]. This integrated workflow typically delivers results within 24–48 h, dramatically accelerating therapeutic decisions compared to weeks required for serological conversion or specialized cultures [34]. Figure 1 illustrates the mNGS technical principles. For deeper discussion on wet-lab workflows and bioinformatic innovations, readers should refer to recent advances in portable on-site sequencing [35, 36], fully automated clinical systems [37], and comparative analyses of mNGS versus targeted NGS platforms [38, 39].

The diagnostic superiority of mNGS manifests in several IE-specific applications, including (i) Detection of fastidious pathogens: mNGS excels in identifying Bartonella spp., Coxiella burnetii, and Tropheryma whipplei, organisms notoriously challenging to cultivate [40]. Recent studies demonstrate mNGS identifying Bartonella henselae directly from excised valves in patients with months of unexplained febrile illness, where blood cultures and serology yielded equivocal results [41]. (ii) Unbiased pathogen agnosticism: Unlike targeted methods, mNGS diagnosed a case of Mycoplasma hominis prosthetic valve endocarditis an organism not considered in initial differentials, highlighting its capacity for unexpected discoveries [42]. (iii) Polymicrobial infection resolution: mNGS revealed mixed infections including Enterococcus faecalis and Granulicatella adiacens infections in a case of culture-negative IE post-cardiac device implantation, fundamentally altering antimicrobial strategy [43]. (iv) Simultaneous DNA/RNA Interrogation: Co-testing allows detection of diverse pathogen types in a single assay, crucial for cases with broad differential diagnoses [44]. (v) Potential for Rapid Turnaround: While current clinical pipelines vary, optimized mNGS workflows can potentially deliver results within 24–48 h of sample receipt, significantly faster than prolonged culture incubation or serological convalescent titers [45]. (vi) Direct Analysis from Complex Samples: mNGS can be applied directly to surgically excised valve tissue (the gold-standard specimen), blood, or emboli, bypassing the need for viable organisms [46]. However, mNGS sensitivity is constrained by low pathogen biomass and high host DNA background, necessitating expert interpretation to differentiate contamination from true infection; semi-quantitative results may obscure minor co-pathogens, and resistance gene detection cannot confirm functional expression or replace phenotypic susceptibility testing [47]. For comprehensive details on mNGS methodologies, laboratory validation [46], diagnostic advantages and limitations, see recent studies by Zhao 2024 [48], Miller 2019 [49], and comparative performance assessments Yi 2025 [32].

Sequencing Platform Evolution and Technical Trade-Offs for IE Applications

The diagnostic performance of mNGS hinges on selecting appropriate sequencing platforms, each with distinct advantages and limitations for IE applications: (i) Short-read platforms (Illumina/BGI) dominate clinical laboratories due to unparalleled accuracy (error rates < 0.1%) and high throughput. The NovaSeq 6000 platform enables deep sequencing of low-biomass samples like blood, enhancing sensitivity for bloodstream infections. However, their read lengths (150–300 bp) struggle with repetitive genomic elements common in resistance islands and virulence cassettes [50]. Recent enhancements in molecular tagging have improved error correction, making these platforms ideal for species-level identification and SNP-based resistance detection 56. (ii) Long-read technologies (Oxford Nanopore) generate reads exceeding 10 kb, enabling complete assembly of resistance operons and virulence gene clusters [51–55]. This proves invaluable when predicting phenotypic resistance patterns in Enterobacteriaceae or Pseudomonas causing prosthetic valve infections. The MinION device portability facilitates rapid on-site sequencing [56], with studies reporting IE diagnosis within 6 h of sample receipt [57, 58]. Despite higher per-base error rates (5–15%), adaptive base-calling algorithms have significantly improved accuracy for clinical isolates [59]. Crucially, Nanopore real-time sequencing permits aborting runs once pathogen-specific signatures are detected, conserving resources a feature leveraged in urgent IE diagnostics. The 2025 ABRF NGS Phase II and other studies provides critical platform comparisons: Illumina platforms achieved 99.4% sensitivity for single-pathogen infections, while Nanopore demonstrated superior performance in resolving mixed infections and structural variants. For IE diagnostics, where both precision and comprehensive resistance profiling matter, a hybrid approach using Illumina for primary detection and Nanopore for resistance/virulence characterization offers a powerful synergistic solution [53, 60–62]. While nanopore sequencing exhibits higher error rates (∼5–15%) compared to Illumina platforms (< 0.1%), its clinical impact varies by use case. For endocarditis diagnostics where the primary goal is pathogen identification to genus/species level rather than fine-scale strain typing, some report suggests accuracy may be sufficient [59].

Diagnostic Interpretation Challenges and Signal Vs. Noise in mNGS Results

mNGS is not a standalone diagnostic solution but a complementary tool within the Duke Criteria framework recommended by major cardiovascular organizations globally (Table 3). Its value lies in enabling pathogen-directed therapy for previously undiagnosed CNE. One of mNGS strengths, its sensitivity (Table 1), is also its major interpretive challenge. Not every sequence read equates to infection. Background flora, environmental DNA, and lab contaminants can all show up in results, especially in low-biomass samples like blood. Salter et al. (2014) showed that even trace contamination from reagents can skew microbiome analyses [63]. In clinical mNGS, the stakes are higher: a false-positive hit can lead to unnecessary treatments or missed diagnoses. That is why modern mNGS protocols rely heavily on contamination controls, blank runs, batch indexing, and UMIs [64]. Despite UMIs and batch controls, reagent-derived contamination persists in 15–30% of blood samples, requiring integration with host biomarkers for reliable interpretation. But beyond technical safeguards, clinical context is paramount. The most powerful mNGS result is one that aligns with patient presentation, imaging, and disease course. In an individual with prosthetic valve dysfunction and persistent fever, the detection of Bartonella henselae (previously undetected by PCR) may be the key to unlocking effective treatment [65]. In a different patient with vague symptoms and low pathogen load, the same result might represent irrelevant noise. Here, multidisciplinary interpretation becomes essential. The best outcomes occur when clinicians, microbiologists, and bioinformaticians interpret results together, triangulating between data, disease, and decision-making [66]. Emerging probabilistic scoring frameworks now weight pathogen likelihood using clinical parameters (e.g., fever duration, epidemiological exposure) and sequence abundance thresholds, reducing false positives in polymicrobial samples. For instance, algorithms like directed network flow entropy (DNFE) integrate host-response biomarkers (e.g., IL-6, procalcitonin) to distinguish pathogens from commensals in real time [67]. Multidisciplinary review teams including microbiologists, infectious disease specialists, and bioinformaticians can enhance diagnostic accuracy by correlating sequencing data with PET/CT localization of paravalvular abscesses or serological profiles. Recent innovations address noise challenges more fundamentally: Roche’s Sequencing by Expansion (SBX) technology improves signal-to-noise ratios via surrogate polymers (Xpandomers) that amplify nucleic acid signals [68], while BGI’s CycloneSeq platform achieves 97% raw read accuracy through asymmetric chemistry and deep learning–based base calling [69]. Nevertheless, standardized clinical reporting frameworks (e.g., ‘Definite/Probable/Contaminant’ classifications) remain essential to align results with therapeutic decisions.

Amplification Strategies and Error Control

Amplification protocols critically influence mNGS performance [82], particularly in endocarditis diagnostics. Whole-genome (WGA) or whole-transcriptome amplification (WTA) is routinely employed in low-biomass samples like blood, enhancing sensitivity by amplifying microbial nucleic acids amidst overwhelming host DNA [25, 83, 84]. However, these methods risk introducing bias through preferential amplification of certain genomic regions and generating chimeric sequences that mimic novel pathogens or introduction of artifact sequence [85–87]. Conversely, amplification-free protocols are preferred for high-biomass specimens (e.g., valve tissue), preserving microbial community structures but requiring deeper sequencing to detect low-abundance targets [25, 64]. For blood samples in CNE, WGA yield higher pathogen detection sensitivity result compared to amplification-free approaches, albeit with reduced specificity due to false positives from reagent contamination [84]. Unique molecular identifiers (UMIs) serve distinct roles based on workflow design. In amplification-dependent protocols, UMIs correct PCR-induced errors by tagging individual molecules pre-amplification [88], enabling bioinformatic consensus calling to eliminate artifactual variants [89, 90]. Their utility diminishes in amplification-free workflows where sequencing errors dominate; here, UMIs offer minimal benefit as they cannot distinguish genuine nucleotide misincorporation during sequencing from true biological variants [88, 91]. Platform-specific error profiles further impact clinical interpretation. Oxford Nanopore higher error rate (∼5–15%) [92], primarily affects base-calling accuracy in homopolymer regions, limiting reliable antimicrobial resistance gene detection. Nevertheless, it suffices for genus/species-level pathogen of 94.% in IE cases [93, 94], making it valuable for rapid intraoperative decisions when valve tissue is available [95]. In contrast, Illumina’s lower error rate supports high-confidence resistance profiling, but delays result by 12–24 h a critical trade-off in unstable patients [96, 97]. However, this accuracy limitation precludes reliable detection of single-nucleotide polymorphisms required for antimicrobial resistance profiling [94, 98], necessitating confirmatory testing when resistance concerns arise.

Therefore, UMIs serve distinct purposes based on workflow design, in amplification-based protocols (most common), they primarily correct PCR errors/duplicates rather than sequencing errors, with studies showing 40–60% artifact reduction [99]. While in amplification-free workflows, UMIs have limited utility for sequencing error correction (Oxford Nanopore R10.4: Q20+; Illumina NovaSeq: Q30+), but may aid in detecting very low-frequency variants [100, 101].

Economic Evaluation and Implementation Equity of mNGS in IE Diagnostics

The Cost of Clarity: Economic and Logistical Barriers To mNGS Adoption

While mNGS may reduce ICU stays, its high upfront cost exacerbates healthcare disparities. Point-of-care sequencing (e.g., Oxford Nanopore) could democratize access but requires validation in low-resource settings. For all its promise, mNGS remains resource-intensive. Per-sample costs range from $500 to $1,000 depending on depth and platform [104], not including data analysis and interpretation overhead. To mitigate these expenses, innovative cost-reduction strategies must be prioritized: reusable flow cells (e.g., Oxford Nanopore) and pooled sequencing runs could substantially lower per-test costs, targeting 40% reductions by 2027. Regional sequencing hubs funded by WHO-Gavi partnerships [105] could further democratize access in resource-limited settings. Compared to traditional diagnostics, blood cultures costing ~$90 or PCR ~$100, mNGS demands a deeper initial investment (illustrated in Fig. 3). But the calculus changes when viewed through a clinical lens. While early targeted therapy may reduce ICU stays in high-risk cohorts (e.g., prosthetic valve IE), cost-efficacy varies by setting, with negligible benefits observed in uncomplicated cases [106–108]. Recent study showed that early mNGS testing improved patient prognosis while reducing diagnostic delays, especially in high-risk populations [109]. Nevertheless, accessibility remains uneven. High-resource centers are leading the way, while lower-resource settings struggle with infrastructure and reimbursement challenges (Fig. 2; Table 4). This disparity risks creating a diagnostic divide in IE care, where some patients benefit from cutting-edge tools while others remain in a trial-and-error loop. Solutions are emerging, low-cost sequencing platforms, centralized analysis hubs, and mobile point-of-care devices, but scalability, regulatory clearance, and clinical validation are still works in progress (Table 5) [76, 94].

Fig. 3.

Comparative cost distribution of mNGS and traditional diagnostic methods. Figure 3 contrasts the cost allocation between mNGS and traditional diagnostic methods. mNGS incurs higher expenses in equipment and data analysis, while traditional methods have greater costs in reagents and labor. The breakdown emphasizes the distinct financial demands of each diagnostic approach

Fig. 2.

Challenges in mNGS implementation for IE. Figure 2 highlights the primary challenges in mNGS adoption: high costs, complex data interpretation, lack of standardized protocols, and data privacy concerns, along with potential solutions like economic analyses, bioinformatics tools, standardization, and enhanced security

Table 4.

Challenges and recommendations for mngs implementation

Challenge	Description	Recommended Solution	Reference
High Cost	High initial costs of equipment and reagents	Development of low-cost sequencing platforms, funding initiatives	[94]
Technical Complexity	Requires specialized training and infrastructure	Training programs, workshops, interdisciplinary collaboration	[66]
Standardization	Lack of standardized protocols and guidelines	Development of consensus guidelines, quality assurance measures	[102]

Table 5.

Summary of validation studies for mngs in diagnosing CNE

Study	Sample Size	Sensitivity	Specificity	Main Findings	Reference
Wu et al. (2024)	195	89.8%	76.3%	High diagnostic yield, reduced false negatives	[110]
Chen et al. (2024)	2325	87%	59%	Comprehensive pathogen detection, improved clinical prognosis	[109]
Diao et al. (2023)	69	95.18%	91.30%	Enhanced detection of rare pathogens, guided targeted therapy	[111]

Standardization and Implementation: Bridging Technical and Clinical Gaps

Contamination risks remain highest in low-biomass samples (e.g., blood); rigorous negative controls are non-negotiable but inconsistently applied. The potential of mNGS in CNE diagnosis is tempered by significant variability across laboratories, where differences in sample handling, sequencing platforms, bioinformatic pipelines, and interpretive criteria yield inconsistent results from identical samples [111, 112]. Standardization spanning nucleic acid extraction to clinical reporting is essential not only for technical consistency but to foster clinical trust [102, 104]. Without harmonized protocols, even high-performing technologies risk irreproducibility in real-world practice (Fig. 2). Adoption of unique molecular identifiers (UMIs), batch decontamination, and clinical-grade bioinformatics pipelines (e.g., VPipe [78]) is critical to reduce variability. Inter-laboratory proficiency testing, such as WHO-led initiatives [63], should validate these protocols globally. Pre-analytical variability remains a critical vulnerability. Inconsistent collection, storage, or processing of blood, tissue, or embolic specimens directly impacts nucleic acid integrity and sequencing depth, leading to false negatives or misclassified pathogens [112]. Analytical challenges further compound this: Illumina platforms offer accuracy but slower turnaround, while Oxford Nanopore enables rapid sequencing at the cost of higher error rates. Platform selection must align with clinical priorities speed for critically ill patients versus depth for polymicrobial infections. The greatest heterogeneity arises in bioinformatics. Custom pipelines vary in host-sequence filtering, pathogen-threshold definitions, and read classification, creating discordant results from the same sample. To mitigate this, adoption of unique molecular identifiers (UMIs), batch decontamination, and WHO-led proficiency testing could reduce variability. Tiered reporting frameworks (e.g., ‘Definite/Probable/Contaminant’) align results with clinical context and threshold-based pathogen scoring is paramount [102, 111]. Clinical-grade bioinformatics tools (e.g., VPipe [78]) and transparent alignment to curated databases reduce false positives and enhance interpretability. Critically, standardized clinical reporting frameworks must evolve alongside technical improvements. A raw pathogen list is inadequate; reports should prioritize actionable insights by: (i) Flagging contaminants and commensals, (ii) Highlighting high-confidence pathogens, (iii) Correlating findings with clinical context (e.g., imaging, serology), (iv) Implementing tiered classifications (e.g., ‘probable pathogen,’ ‘uncertain significance’) to guide therapeutic decisions [98, 111]. Inter-laboratory proficiency testing (e.g., WHO-led initiatives [63] and regulatory collaboration (FDA/EMA) are vital for validation. Until universal standards exist, clinicians should interpret mNGS within a multidisciplinary framework integrating microbiology, radiology, and patient evolution to navigate ambiguity.

Trends and Gaps in Current Research

Emerging Trends

The clinical integration of mNGS has gained remarkable traction due to its capability to detect a broad spectrum of pathogens with higher sensitivity than conventional methods (Fig. 4; Table 1). Studies have demonstrated its versatility across a variety of infectious diseases, including CNE [66, 104]. These investigations report diagnostic yields ranging from 80% to over 90%, particularly in individuals with prior antibiotics treatment. This variability stems from pre-analytical factors (e.g., sample degradation during transport in low/middle-income countries (LMICs)) and analytical disparities (e.g., Illumina vs. Nanopore error rates), highlighting irreproducibility risks in real-world practice. However, this high sensitivity is not universal. Previous studies reported more modest yields (60–75%) in mixed cohorts, highlighting the influence of sample type, pathogen load, and sequencing depth [112]. Such discrepancies reveal the necessity for further refined protocols that account for contextual variables. In terms of bioinformatics, significant progress has been made with machine learning-driven pipelines that improve signal-to-noise ratios and pathogen classification. One study emphasized these advances but also cautioned against overreliance on unvalidated algorithms, particularly in low-biomass samples prone to contamination [102]. Furthermore, another advocated for AI-assisted interpretation tools that integrate clinical metadata to improve specificity, critics argue that these systems are only as robust as the datasets they are trained on, which are often limited in size and diversity [76]. As such, while AI shows promise, it should currently serve as a decision-support mechanism rather than a replacement for expert review.

Fig. 4.

Integration of mNGS into clinical practice. Figure 4 depicts the integration of mNGS in diagnosing CNE. It outlines the process from sample collection, library preparation, and sequencing, through data analysis and result interpretation, to reporting and clinical decision-making, emphasizing the roles of mNGS in identifying difficult-to-detect pathogens and guiding precise treatment

Multimodal Diagnostic Integration and Clinical Decision Pathway

The integration of mNGS with advanced diagnostic modalities particularly advanced imaging and serological biomarkers can significantly enhances diagnostic precision in CNE (Table 6). For prosthetic valve endocarditis (PVE), combining mNGS with ¹⁸F-FDG PET/CT could improve infectious focus localization in cases with ambiguous clinical signs [113, 114]. This synergy could arises from PET/CT capacity to identify metabolically active inflammation (SUVmax > 4.1) and can guide targeted mNGS sampling of resected valve tissue, thereby reducing false positives from environmental contaminants [113, 115]. Conversely, in seropositive Bartonella IE (IgG titers > 1:1,024) [116, 117], mNGS could add marginal diagnostic value but remains critical for detecting co-infections or resistance markers missed by serology alone [118]. Emerging molecular imaging probes might further augment mNGS utility. Antimicrobial peptide-based radiotracers (e.g., ⁹⁹ᵐTc-UBI) selectively bind bacterial cell walls, which might enable real-time visualization of Staphylococcus aureus biofilms on prosthetic valves [113]. When correlated with mNGS-derived pathogen data, these probes can differentiate active infection from sterile inflammation a longstanding challenge in IE management [113, 119]. Similarly, host-response biomarkers like procalcitonin and IL-6 might refine mNGS interpretation: elevated levels (> 2 ng/mL and > 40 pg/mL, respectively) strengthen the clinical relevance of low-abundance mNGS reads by confirming systemic inflammation [120, 121]. However, multimodal integration faces interoperability barriers. Current electronic health records rarely consolidate genomic, imaging, and biomarker data into unified dashboards, delaying time-sensitive therapeutic decisions. This might be address by embedding mNGS within a multi-omics framework that can tracks pathogen dynamics alongside host immune responses in IE patients, aiming to establish causality between specific microbial profiles and thrombotic complications. Preliminary data reveal that Streptococcus sanguinis detected by mNGS correlates with colony vegetations on CT enterography, prompting colonoscopy that identified occult malignancies in few cases [115]. Diagnostic workflows must be context-sensitive to maximize cost-efficacy. For high-probability IE (e.g., febrile injection drug users with tricuspid vegetations), echocardiography plus blood culture suffices, reserving mNGS for treatment failures [118]. Contrastingly, indeterminate cases (e.g., culture-negative PVE with nonspecific CRP elevation) warrant concurrent mNGS and PET/CT as first-line tools [113, 114]. This stratification could optimizes resources use: Biofire BJP panels automated PCR assays detecting 31 pathogens and 10 resistance genes in 1 h provide rapid screening, while 16 S/18S rDNA PCR covers rare pathogens like Lactococcus garvieae albeit with longer turnaround (100 h) [118]. Validated frameworks (e.g., ESC 2023 guidelines [122]) now could recommend tiered algorithms where BJP-negative samples trigger reflex mNGS testing [114, 118]. However, multimodal integration faces substantial interoperability challenges. Harmonizing mNGS data with electronic health records remains nascent, and cross-disciplinary workflows (e.g., correlating sequencing results with real-time echocardiographic findings) require specialized platforms unavailable in resource-limited settings [114, 123]. The 2025 Swiss IE cohort study further revealed socioeconomic disparities: high-income centers utilized integrated diagnostics in > 85% of complex cases versus < 15% in regions lacking PET/CT infrastructure [124]. Future efforts must prioritize unified data architectures that embed mNGS within endocarditis team workflows, enabling dynamic reinterpretation of results alongside advancing clinical parameters. For comprehensive detail, see Zeng et al.’s (2023) work for synchronizing blood mNGS with surgical findings, the CIVA consortium guidelines for imaging-guided tissue sampling [115, 123], and the FISHseq protocol for prosthetic valve biofilm characterization [119].

Table 6.

Proposed IE diagnosis and treatment strategies for future care

Proposed Strategy	Diagnostic/Treatment Innovation	Potential Benefits	Implementation Considerations	References
Integrating AI and Machine Learning	AI-driven algorithms to interpret mNGS data and imaging results	Increased diagnostic accuracy, reduced human error, faster data processing	Requires extensive training datasets, regulatory approval, and cross-disciplinary expertise	[72, 73]
Personalized Medicine Approach	Utilizing genomics and mNGS data to tailor antibiotic therapy	Improved treatment efficacy, reduced side effects, optimized antibiotic use	High costs, need for robust bioinformatics infrastructure, patient consent and data privacy concerns	[74, 75]
Point-of-Care mNGS Devices	Development of portable, rapid mNGS devices for bedside use	Immediate pathogen detection, faster clinical decision-making, improved patient outcomes	Developmental and validation challenges, integration into clinical workflows, cost considerations	[76]
Telemedicine and Remote Monitoring	Remote monitoring of patients using digital health tools and mNGS results	Enhanced patient follow-up, early detection of complications, reduced hospital readmissions	Data security, infrastructure requirements, patient and provider training	[77]
Advanced Bioinformatics Platforms	Enhanced platforms for real-time data analysis and pathogen identification	Greater diagnostic precision, streamlined workflow, integration with electronic health records (EHRs)	Requires significant investment, interoperability with existing systems, continuous updates	[78]
Next-Generation Antibiotics	Development of new antibiotics targeting resistant pathogens identified by mNGS	Effective treatment of antibiotic-resistant infections, reduced mortality and morbidity	Long drug development timelines, high costs, regulatory hurdles	[79]
Microbiome-Based Diagnostics	Leveraging microbiome profiling to understand pathogen interactions and diagnose IE	Insights into pathogen communities, potential for targeted therapies, reduced false positives	Complexity of microbiome data, need for advanced analytical methods, clinical validation	[80, 81]

Advancing Methodological Innovations and Clinical Validation

The integration of artificial intelligence (AI) with mNGS could advances the diagnostic precision for CNE (Table 6) [125]. Recent multimodal AI frameworks, such as the MAARS model for arrhythmic death risk stratification in hypertrophic cardiomyopathy, demonstrate how combining electronic health records, imaging data, and genomic insights can achieve AUCs of 0.89–0.81 significantly outperforming traditional clinical guidelines (AUC 0.22–0.35) [126]. Similarly, the EchoNext system leverages deep learning to detect structural heart disease (e.g., low LVEF, valve pathology) from routine electrocardiograms with 85.2% AUROC, enabling targeted use of echocardiography in high-risk cohorts [127]. These advances highlight AI capacity to extract latent diagnostic signals from conventional tests, though their applicability to mNGS-based pathogen detection requires further validation summarized in Table 5 [128, 129]. Hybrid AI architectures now merge convolutional neural networks (CNNs) for spatial pattern recognition with large language models (LLMs) that interpret unstructured clinical notes [130, 131]. Such frameworks achieve 95.1% accuracy in heart failure prediction by synthesizing demographic, electrocardiographic, and textual data [132]. For mNGS, analogous models could contextualize sequencing results with patient history and imaging findings, distinguishing pathogens from contaminants more effectively. For instance, lipid-rich necrotic cores or positive remodeling on coronary CT angiography features associated with vulnerable plaques may correlate with specific microbial profiles in IE [133]. However, AI performance remains heterogeneous across populations. While EchoNext maintains AUROC > 78% across diverse healthcare settings [127], models like CardioMind (China’s first cardiovascular AI) exhibit geographic bias, with training data skewed toward Eastern cohorts. This underscores the need for globally representative datasets to ensure equitable generalizability. Critical implementation barriers persist. First, explainability: Even high-performing ‘black-box’ models like MAARS require Shapley Additive exPlanations (SHAP) to clarify decision pathways and foster clinician trust [132]. Second, data quality: AI-mNGS integration depends on standardized preprocessing; variable nucleic acid extraction methods alter pathogen recovery by > 25%, confounding algorithmic predictions [133]. Third, clinical workflow integration: While EchoNext reduced unnecessary echocardiograms by 41% in prospective trials [127], no studies yet demonstrate AI-guided mNGS improving IE outcomes in real-world settings. Future efforts must prioritize prospective validation in multicenter cohorts, emphasizing hard endpoints like mortality or embolic events rather than analytical performance alone. For methodological advancements, see the hybrid CNN-LLM framework by Pinna et al. (2025) and plaque characterization AI in coronary imaging by Alshraideh et al. (2025) [132, 133].

Identified Research Gaps

Despite promising data, several critical gaps persist. First, the absence of large, multicenter validation studies undermines generalizability (Table 5). While one study demonstrated robust results in a high-volume center [104], another report highlighted that, smaller hospitals with limited bioinformatics support struggled to replicate those findings [65]. This indicates that mNGS performance is not solely a function of the technology, but also of institutional capability. Second, there is a marked lack of consensus on interpretive thresholds. The mNGS sensitivity (Table 1), can yield ambiguous results, particularly in distinguishing pathogens from contaminants or commensals. For example, a study reported several false positives due to inadequate filtering [112], while a study implemented strict confidence scoring to mitigate this issue [111]. These divergent approaches complicate inter-study comparison and point to the critical needs for standardized reporting criteria. Cost-effectiveness remains another unresolved issue (Table 7). While one study reported that early mNGS testing reduced ICU stay and antibiotic use [109], a contrasting analysis by another study found no significant difference in overall treatment costs when controlling for illness severity [134]. mNGS offers higher diagnostic yields than PCR and blood cultures but at substantially greater upfront cost summarized in Table 7. More comprehensive economic modeling is needed, accounting for diagnostic performance and long-term outcomes such as recurrence rates, mortality, and antimicrobial resistance. Finally, long-term outcome data are sparse. While several studies show immediate benefits in terms of diagnosis and therapeutic adjustment, few extend follow-up beyond hospital discharge. Additionally, > 80% of cited studies originate from high-income regions, potentially overestimating mNGS efficacy in LMICs where cold-chain logistics and electrical instability degrade sample integrity. Without longitudinal data, it remains unclear whether mNGS-guided care improves survival, reduces recurrence, or enhances quality of life. Future research should prioritize patient-centered endpoints to truly define clinical utility. These trends and contrasting findings underscore a field in evolution where mNGS has established its potential but must now earn its place in standardized clinical care. The following section explores how innovation, regulation, and strategic investment must converge to unlock mNGS full promise as a globally accessible diagnostic standard.

Table 7.

Cost-benefit analysis of diagnostic methods

Diagnostic Method	Cost per Sample ($)	Time to Result	Clinical Accuracy	Cost-Effectiveness	Reference
Traditional Blood Culture	84–97	Days to Weeks	Moderate	Low	[135, 136]
PCR	50–200	Hours to Days	High	Moderate	[137]
mNGS	1000–2500	24–48 Hours	Very High	High (long-term)	[106]

Future Directions and Recommendations

The evolution of mNGS for CNE demands urgent and coordinated action, three key technological innovations must be prioritized. Emerging hybrid architectures theoretically enable rapid, high-accuracy sequencing, though clinical validation remains pending [84, 138]. Concurrently, integrated host-pathogen cartridges combining mNGS with immune biomarkers (e.g., IL-6, procalcitonin) [64, 139] using microfluidic technology would standardize contamination-prone extraction steps while providing immunological context to differentiate pathogens from commensals. Finally, WHO-coordinated open-source AI platforms sharing globally validated algorithms must be deployed, with mandatory bias audits ensuring > 30% representation of LMIC populations in training datasets to prevent algorithmic disparities [63, 140, 141]. For equitable implementation, health systems must address cost barriers challenges (Table 4) through manufacturer-subsidized flow-cell reuse and pooled sequencing runs targeting 40% cost reduction by 2027, complemented by WHO/Gavi-funded regional sequencing hubs to serve LMIC networks [105]. Reimbursement models should directly link mNGS coverage to documented clinical benefits: insurers could require proof of ≥ 48-hour reductions in broad-spectrum antibiotic use or prevention of ≥ 1 embolic event through early pathogen-directed therapy [142, 143]. Critical research gaps demand immediate attention. Prospective trials comparing mNGS-guided versus conventional CNE management are urgently needed [109, 134]; an international registry tracking 30-day mortality/recurrence across 5,000 + patients would provide definitive evidence. Multicenter studies must also validate pathogen-specific host-response thresholds (e.g., Bartonella-induced IFN-γ cutoffs) to resolve diagnostic ambiguity [144, 145], while simplified workflows using freeze-dried reagents and smartphone-based bioinformatics require validation in LMIC field trials to enable low-infrastructure deployment [105, 146]. A tripartite call to action is essential, clinicians should establish multidisciplinary rapid-response teams implementing tiered reporting frameworks aligned with modified Duke criteria [4, 147]. Regulators must fast-track accreditation for integrated assays by 2026 and mandate biannual proficiency testing using CNE reference panels [148]. Policymakers ought to fund mNGS inclusion in WHO Essential Diagnostics Lists and tie hospital reimbursements to diagnostic yield audits (> 85% pathogen identification in CNE). Without these measures, mNGS risks exacerbating global disparities where early pathogen-directed therapy becomes standard in high-income settings while empirical delays persist in regions bearing the highest CNE burden. The 18–25% mortality of untreated CNE [1, 2] demands we bridge this gap urgently. By uniting innovation with equity, we can transform mNGS from a privileged technology into a cornerstone of precision global health.

Conclusion

mNGS significantly advances pathogen detection in CNE, identifying causative organisms in > 90% of cases where conventional diagnostics fail, However, it faces cost barriers, standardization gaps, and interpretative challenges that require resolution before routine adoption (Table 4). The mNGS future in CNE diagnosis depends on its transformation from a high-potential technology into a robust, accessible, and regulated diagnostic system. Key constraints include false positives in low-biomass samples, inter-lab variability, and exclusion of 37% of eligible studies from low-resource settings in analysis highlighting geographic evidence gaps. As the field advances, success will be measured not by sequencing speed or depth alone, but by meaningful improvements in patient care, diagnostic equity, and system-level efficiency. Achieving this vision requires a concerted effort across science, policy, and practice, linking innovation with infrastructure, evidence with education, and discovery with delivery. Additionally, the high costs and technical complexity of mNGS, coupled with the limitation in standardization, hinder widespread adoption (Table 4). Overcoming these barriers will require ongoing research, greater interdisciplinary collaboration, and innovative strategies such as the incorporation of artificial intelligence for data analysis, development of portable mNGS devices for point-of-care use, and the integration of multi-omics approaches. Its clinical adoption requires cost reduction, interoperability with electronic health records, and guidelines defining actionable results. These advancements, along with improved bioinformatics tools and telemedicine for continuous monitoring, hold promise to enhance IE diagnostics and treatment, ultimately enhancing patient care in culture-negative cases.

Abbreviations

mNGS

Metagenomic Next-Generation Sequencing

CNE

Culture-Negative Endocarditis

IE

Infective Endocarditis

PCR

Polymerase Chain Reaction

16S rRNA

16 S Ribosomal Ribonucleic Acid

AI

Artificial Intelligence

QUADAS-2

Quality Assessment of Diagnostic Accuracy Studies-2

ESC

European Society of Cardiology

AHA

American Heart Association

ICU

Intensive Care Unit

PET/CT

Positron Emission Tomography/Computed Tomography

DNA

Deoxyribonucleic Acid

RNA

Ribonucleic Acid

WHO

World Health Organization

FDA

U.S. Food and Drug Administration

EMA

European Medicines Agency

Author Contributions

SAUS, BZ; Conceptualization, investigation, methodology, resources, validation, writing – original draft, writing – review & editing, YQ & BXZ: Investigation, methodology, resources, HMA& YTVM: literature search validation, software and visualization, GBF & KZ: Project administration, supervision, validation, visualization, writing – review & editing, LZ & MG: Investigation, methodology, resources, supervision and validation, YY: Conceptualization, funding acquisition, project administration, supervision, validation, writing – review & editing. All authors have read and agreed to the submission of the manuscript to Journal of Epidemiology and Global Health.

Funding

This work was supported by Sun Yat-sen University Eighth Hospital Cohort Construction Project with the No. (GCCRCYJ065).

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Ethics Approval and Consent to Participate

This study does not involve interaction with animal or human subjects or the collection of personal information, no ethics committee/IRB approval was required for this research, in accordance with the Helsinki Declaration.

Consent for Publication

Not applicable.

Competing Interests

The authors declare no competing interests.