Comparison of the microbiome of sputum from HIV/AIDS patients using PacBio 16S rRNA sequencing and ddPCR

Abstract

Background

Pulmonary infections are prevalent opportunistic complications among HIV/AIDS patients, necessitating precise pathogen identification for targeted treatment. The traditional culture methods used for diagnosing these infections have inherent limitations when it comes to detecting the diverse microbial flora present in HIV/AIDS patients.

Methods

In this study, sputum samples were collected from 37 HIV/AIDS patients admitted to Mengchao Hepatobiliary Hospital of Fujian Medical University. A total of 45 samples were subjected to analysis using three different methods: traditional culture methods, third-generation long-read sequencing (PacBio), and droplet digital PCR (ddPCR). The study aimed to compare the diagnostic capabilities and accuracy of microorganism identification among these methods.

Results

Out of 45 sputum samples, the traditional culture method identified only 5 positive cases, whereas PacBio sequencing detected all samples positively. Among the 7 samples (including 5 traditional culture positive and 2 negative samples), PacBio sequencing was able to identify 51 species (77 strains), while ddPCR detected 15 species (64 strains). In contrast, traditional culture methods were only able to identify 3 bacterial species (6 strains). In samples that tested positive by culture, both PacBio sequencing and ddPCR were consistent in identifying the primary microorganisms, indicating the presence of multiple concurrent infections. In samples that were negative by culture, both PacBio and ddPCR were able to detect a broader range of microorganisms, showing a higher sensitivity compared to traditional methods. PacBio sequencing provided a comprehensive view of the microbial flora, whereas ddPCR enabled rapid and precise detection of common microorganisms and antibiotic resistance genes, compensating for PacBio’s shortcomings in absolute quantification.

Conclusions

The study demonstrates that PacBio sequencing and ddPCR offer significant advantages over traditional culture methods in diagnosing pulmonary infections in HIV/AIDS patients. These advanced techniques not only detect a greater variety of pathogens but also provide a more detailed understanding of the microbial flora and resistance profiles, which is crucial for effective clinical management. The combination of PacBio’s comprehensive profiling and ddPCR’s rapid detection capabilities presents a powerful approach to overcoming the limitations of traditional diagnostic methods in this patient population.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-025-11500-6.

Background

Pulmonary infections present significant risks to elderly and immunocompromised individuals, especially those with AIDS. Timely and accurate diagnosis is essential for effective disease management and optimal outcomes. In these patients, the range of potential pathogens expands to include various bacteria and fungi. Early pathogen detection supports effective antibiotic stewardship, shortens hospital stays, and improves survival rates. Delayed diagnosis may impede targeted therapies, resulting in suboptimal treatment and worse prognoses [1].

AIDS patients are highly vulnerable to opportunistic infections (OIs) due to severe immunodeficiency resulting from HIV-induced destruction of CD4⁺T lymphocytes. As the disease progresses, CD4⁺T lymphocyte counts typically fall below 200 cells/µL, significantly increasing the risk of lung infections. These infections often present with severe symptoms, rapid progression, non-specific manifestations, and atypical imaging findings, posing diagnostic challenges [2–4]. Respiratory failure is common in severe lung infections in AIDS patients, contributing to high mortality rates [5]. Traditional culture methods, while providing morphological and growth characteristics of pathogens, are limited by long culturing times and reduced sensitivity for detecting certain pathogens.

Recent advancements in sequencing technologies have significantly improved our ability to study the Human microbiome. Full-length 16 S rRNA gene sequencing using PacBio technology offers higher taxonomic resolution compared to traditional methods, as demonstrated in a recent study by Jordán-López et al. [6]. In this study, we compare the performance of PacBio 16 S rRNA sequencing and ddPCR in analyzing the microbiome of sputum samples from HIV/AIDS patients. This comparison aims to provide insights into the strengths and limitations of each method in identifying pathogenic microorganisms.

PacBio SMRT sequencing uses the Single-molecule real-time (SMRT) chip for sequencing during synthesis [7], achieving up to 99.999% accuracy with its circular consensus sequencing (CCS) mode, surpassing Nanopore sequencing [8]. This technology is especially effective for microbial species identification via the 16 S rRNA gene sequence (~ 1.5 kb) [9]. In contrast, metagenomic next-generation sequencing (mNGS), such as Illumina and UW BGI sequencers, is limited to a maximum read length of 300 bp, allowing only genus-level identification and prone to splicing errors [10]. Third-generation sequencing with long reads (> 1 Mb) enables microbial macro-genome sequencing and direct full-length 16 S rRNA measurement, achieving 99.999% accuracy at 30X sequencing depth. PacBio sequencing for 16 S rRNA gene sequencing offers cost-effectiveness, high throughput, completeness, and accuracy [11, 12].

The advantages of PacBio SMRT sequencing include ultra-long read length, enabling full-length 16 S rRNA gene sequencing [13]; high accuracy for reliable analysis [14]; uniform base coverage without GC content bias; and direct detection of base modifications using color-tagged dNTPs [15].

In 1992, Sykes et al. [16] laid the foundation for the development of droplet digital PCR (ddPCR) technology by employing limited dilution of samples to ensure that each microwell contained only a single template molecule. This approach aimed to accurately determine the number of original molecules by calculating the amplification signals after PCR, particularly when detecting low-abundance IgH heavy-chain mutated genes in complex backgrounds. In 1999, Vogelstein and Kinzler [17] first proposed the concept of ddPCR. To measure K-RAS mutations in colorectal cancer patients, they performed limited dilution of samples and distributed them into a 384-well plate for reaction, successfully detecting K-RAS mutations. However, the manual operation required for sample distribution limited the uniformity and quantity of nucleic acid allocation, thereby restricting the development of ddPCR technology. In recent years, advancements in microfluidics, water-in-oil emulsion droplet technology, and nanofabrication have overcome these technical bottlenecks. These innovations have enabled the development of high-throughput, automated, and commercial ddPCR platforms. Consequently, ddPCR technology has been rapidly adopted in clinical applications [18].

The ddPCR system partitions a fluorescent PCR reaction into nanoscale microdroplets, each acting as an independent PCR reactor. After amplification, the fluorescence signal is digitized, enabling precise quantification of specific microorganisms based on Poisson distribution principles. ddPCR is particularly effective for detecting target DNA at low concentrations, in degraded samples, or with limited materials, and is widely used for gene copy number variations, gene expression quantification, tumor marker detection, and infectious disease biomarker identification [19–21]. ddPCR also plays a significant role in the diagnosis of viral infections and has been applied in the detection of various viruses, including human herpesvirus, cytomegalovirus, influenza virus, hepatitis B virus, and HIV [22]. The technology has been included in the “Expert Consensus on Clinical Application of Nucleic Acid Testing for Pathogen Diagnosis in Adult Respiratory Infections (2023)” [23]. The consensus indicates that ddPCR can directly quantify the number of pathogens in respiratory specimens. It is suitable for detecting low-concentration samples, low-abundance genes, or mutated genes. Additionally, ddPCR can be used to determine the detection limits of other nucleic acid testing methods and serves as a reference for clinical practice.

With the reduction in sequencing costs, sequencing technology has become more prevalent in medicine. By leveraging high-throughput sequencing and ddPCR, the traditional microbial culture process can be bypassed, addressing challenges such as delays, low positivity rates, and difficulties in confirming pathogenicity. These techniques also overcome the limitation that in vitro cultures may not accurately reflect in vivo infections. High-throughput sequencing can identify wide range of species and quantify their proportions, while ddPCR excels in the absolute quantification of low concentrations of pathogens. ddPCR is especially useful for the rapid detection of common pathogens and antibiotic resistance genes in lung infections, compensating for high-throughput sequencing’s limitations in quantification. Conversely, high-throughput sequencing is better suited for diagnosing infections caused by classical or unrecognized pathogens [24], offering a comprehensive view of pathogens in HIV co-infected lungs.

Recent studies, including the Human Microbiome Project (HMP) [25], have provided extensive data on the microbiome of the oral and pharyngeal regions in healthy individuals. These studies have identified a diverse range of microbial communities and their potential roles in maintaining health. However, making meaningful comparisons between these healthy individuals and those with HIV/AIDS is challenging due to high inter-individual variability. Factors such as age, diet, hygiene practices, and genetic background can significantly influence the microbiome composition. Additionally, a recent longitudinal study by Bach et al. [26] demonstrated significant temporal variations in the oropharyngeal microbiota, further complicating the comparison process.

This study aims to evaluate the diagnostic performance of PacBio sequencing, ddPCR, and conventional methods in detecting pathogens in sputum samples from HIV/AIDS co-infected patients, by comparing their efficiencies.

Methods

Sample collection

This study involved the collection of sputum specimens from 45 patients diagnosed with AIDS and concurrent pulmonary infections at Mengchao Hepatobiliary Hospital of Fujian Medical University. None of the patients in our study cohort experienced bacteremia during the study period. Ethical approval was granted by the Ethics Committee of Mengchao Hepatobiliary Hospital, Fujian Medical University (Fuzhou, China; Ethics No. 2020-088-01).

The diagnostic criteria for AIDS followed the Third Edition of the AIDS Diagnostic and Treatment Guidelines (2021) [4]. The diagnostic criteria for lung infection were based on the Diagnostics (9th edition), published by the People’s Medical Publishing House [27]. The diagnostic criteria for opportunistic infections were based on Guidelines for prevention and treatment of opportunistic infections in HIV-infected adults and adolescents (NIH, CDC, HVMA, IDSA) [28]. Patients aged ≥ 18 years, diagnosed with AIDS and concurrent lung infections, were included in the study. Sputum specimens were collected following the standard operating procedures outlined in the National Guide to Clinical Laboratory Procedures (4th edition), published by the People’s Medical Publishing House [29]. The sputum samples were divided into two aliquots: one for 16 S rRNA sequencing and ddPCR analysis, and the other for traditional microbial culture and identification.

All HIV/AIDS participants were recruited from Mengchao Hepatobiliary Hospital of Fujian Medical University—the former Fuzhou Infectious Disease Hospital and the provincially designated referral centre for HIV/AIDS care in Fujian Province, China.

Limitations and classification of microorganisms

Our study did not include control samples from healthy individuals, which is a significant limitation. Without a healthy control group, it is challenging to definitively classify microorganisms as strictly commensal or pathogenic based solely on our data. In the context of our study population, which consists of HIV/AIDS patients, the distinction between commensal and pathogenic microorganisms is particularly nuanced. Immunocompromised individuals are at a higher risk of opportunistic infections, and microorganisms that are typically considered commensal in healthy individuals can become pathogenic due to their weakened immune systems. This makes it difficult to draw a clear line between commensal and pathogenic microorganisms in this specific patient population.

Traditional sputum culture and pathogen identification

Sputum samples were inoculated onto blood and chocolate agar plates (Autobio Diagnostics, China) for bacterial culture. The plates were then incubated at 37℃ in a 5% CO₂ atmosphere for 1–3 days, which is a standard incubation period commonly used in clinical microbiology laboratories to detect the majority of clinically relevant bacterial pathogens. This incubation period is designed to provide timely results, which is critical for rapid diagnosis and treatment in clinical settings. For each type of suspected species, three to five representative colonies were selected from the blood or chocolate agar plates based on their morphological characteristics. Initial differentiation was performed using standard biochemical tests. Then, the cultured pathogenic microorganisms were subsequently identified using VITEK-MS (BioMerieux SA, BioMerieux Inc., France).

Pre-treatment of sputum samples and DNA extraction

Sputum was liquefied for 30 min with a 4% NaOH solution. After centrifugation at 12,000×g for 5 min, the supernatant was discarded, and the pellet was resuspended in 2 mL of 1× PBS buffer. A second centrifugation at 12,000×g for 5 min was performed, the supernatant was removed, and the pellet was finally resuspended in 1× PBS buffer. DNA was extracted from the sputum samples using the STE method (Tris-base, NaCl, EDTA) [30].

Third-generation PacBio full-length 16 S rRNA sequencing

The samples were sent to Biomarker Technologies (Beijing, China) for PacBio Sequel sequencing. The turnaround time for the sequencing run, including library preparation, sequencing, and bioinformatics analysis, was approximately one month. The samples were collected over an extended period, but for individual samples, the process from collection to DNA extraction can be completed within a day. To optimize the use of sequencing resources and reduce costs, the 45 samples were batched together for library preparation and sequencing on a single PacBio flow cell. The full-length 16 S rRNA gene were amplified with primer pairs 27 F: AGRGTTTGATYNTGGCTCAG and 1492R: TASGGHTACCTTGTTASGACTT. Both the forward and reverse 16 S primers were tailed with sample-specific PacBio barcode sequences to allow for multiplexed sequencing. We chose to use barcoded primers because this reduces chimera formation as compared to the alternative protocol in which primers are added in a second PCR reaction. The KOD One PCR Master Mix (TOYOBO Life Science) was used to perform 25 cycles of PCR amplification, with initial denaturation at 95 °C for 2 min, followed by 25 cycles of denaturation at 98 °C for 10 s, annealing at 55 °C for 30 s, and extension at 72 °C for 1 min 30 s, and a final step at 72 °C for 2 min. The total of PCR amplicons was purified with VAHTSTM DNA Clean Beads (Vazyme, Nanjing, China) and quantified using the Qubit dsDNA HS Assay Kit and Qubit 3.0 Fluorometer (Invitrogen, Thermo Fisher Scientific, Oregon, USA). After the individual quantification step, amplicons were pooled in equal amounts. SMRTbell libraries were prepared from the amplified DNA by SMRTbell Express Template Prep Kit 2.0 according to the manufacturer’s instructions (Pacific Biosciences). Purified SMRTbell libraries from the pooled and barcoded samples were sequenced on a PacBio Sequel II platform (Beijing Biomarker Technologies Co., Ltd., Beijing, China) using Sequel II binding kit 2.0. (Please adjust the primer according to the actual amplification sub region).

To ensure sufficient sequencing depth for the 16 S gene of the 45 samples, we employed a pooling strategy and worked closely with the sequencing service provider. After sequencing, we conducted a detailed statistical analysis of the sequencing data for each sample. The results showed that each sample generated an average of 30,000 CCS (Consensus Circular Sequences), which is adequate for accurately analyzing the microbial community structure. To further validate the reliability of the sequencing results, we employed multiple bioinformatics methods for quality control and filtering, including the removal of low-quality sequences and chimeric sequences. Additionally, we compared the sequencing data against known microbial databases to accurately identify the types and relative abundances of microorganisms in the samples. These measures ensured that the sequencing data accurately reflected the actual microbial community structure of the samples.

The downstream data were processed using SMRTlink analysis software, which generated CCS files and converted them into the FASTQ format. Comprehensive 16 S rDNA analysis was performed using BMKCloud, which included long read sequence extraction, barcode recognition, CSS length filtering, and chimera removal to produce high-quality sequences. Clean reads then were conducted on feature classification to output an ASVs (amplicon sequence variants) by DADA2 (Version 1.14.1) [31], and the ASVs conuts less than 2 in all samples were filtered. Taxonomy annotation of the ASVs was performed based on the Naive Bayes classifier in QIIME2 (Version 2021.4) [32] using the SILVA database [33] (release 138.1) with a confidence threshold of 70%. The Alpha diversity was calculated and displayed by the QIIME2 and R software, respectively.

For the analysis of microbial community composition, we used the SILVA database (release 138.1) as the reference database [33]. A simple Bayesian classifier, combined with a comparative method, was applied to assign taxonomic labels to feature sequences. A multi-level statistical analysis of microbial community composition across kingdoms, orders, phyla, families, genera, and species in individual samples was then performed. QIIME2 software was used to compile species abundance tables across different taxonomic levels [32]. The species composition bar chart at the species level was generated using the R programming language (Version 4.3.1) within the RStudio IDE (Version 2025.05.0), with the ggplot2 package (Version 3.5.0) for core visualization and the ggrepel package (Version 0.9.6) to enhance label clarity in the bar chart. In our analysis, we considered species with a relative abundance of at least 1% within a sample as significant for inclusion. This threshold helps to filter out potential noise and focus on the most relevant species contributing to the microbiota profile. The remaining species were grouped under “Others.” Additionally, “Unclassified” refers to species that lack taxonomic annotations.

Droplet digital PCR

Pathogens and AMR genes were detected using six assay panels (four for bacteria, one for fungi, and one for AMR genes) on a five-channel fluorescence ddPCR system. The reagent kit used was the Twenty Bloodstream Infection Pathogens and AMR Genes Nucleic Acid Detection Kit by Digital PCR, manufactured by Pilot Gene Technology Company (Hangzhou, China). The kit model is MP02, with catalog number 8,000,004. This kit has obtained CE-IVD certification and is designed for use with the D3200 Digital PCR System. The overall sensitivity of the multiplex assay is 84.9%, and the specificity is 92.5%. The pathogens or AMR genes targeted in this study are listed in Table 1.

Table 1.

Fluorescence channels and detection targets of DdPCR

Panel	FAM	VIC	ROX	CY5	CY5.5
P1	Pseudomonas aeruginosa	Escherichia coli	Klebsiella pneumoniae	Acinetobacter baumannii	/
P2	Staphylococcus aureus	Candida species	Enterococcus species	Streptococcus species	/
P3	Stenotrophomonas maltophilia	Enterobacter cloacae	Proteus mirabilis	Coagulase-negative staphylococci, CNS	Serratia marcescens
P4	Salmonella	Citrobacter species	Burkholderia cepacia	Bacteroides fragilis	Morganella morganii
AMR-1	KPC	mecA	OXA-48	NDM/IMP	VanA/vanM

ddPCR analysis was performed using the Pilot Gene Droplet Digital PCR System, following the manufacturer’s protocol. A combined detection kit for 25 pathogenic microbial genera and drug-resistant gene nucleic acids, used for the diagnosis of bloodstream infections, was employed (Pilot Gene Technologies, Hangzhou, China). The ddPCR premix for each assay consisted of 3 µL of 5× ddPCR premix, 1 µL of forward primer (10 µM), 1 µL of reverse primer (10 µM), 2 µL of probes (300 nM), 5 µL of isolated sputum DNA, and 3 µL of DNase-free water, for a total volume of 15 µL. ddPCR was performed using a Chip Scanner (CS5) to analyze the droplets for 35 min. Each well of the ddPCR system generates approximately 20,000 to 25,000 droplets per run. For each ddPCR run, both positive and negative controls were included. Synthetic DNA fragments were used as positive controls for each target to monitor the assay performance. DNase-free water was employed as a negative control to eliminate the possibility of external or reagent microbial contamination. Additionally, a standard positive control was included with each new batch of reagent kits during their first use to exclude false negatives. Data analysis for droplet counting and amplitude was performed manually for 30 min using GenePMS software. DNase-free water was used as a negative control to monitor external or reagent microbial contamination and cross-sample contamination.

Results

This study included 45 sputum samples from 37 AIDS patients, including duplicates from eight patients collected at least one month apart.

Detection of pathogenic bacteria by traditional culture methods

Using traditional method, three bacterial species were identified, comprising six strains, all classified as Gram-negative. These included Escherichia coli (3 strains), Pseudomonas aeruginosa (2 strains), and Klebsiella pneumoniae (1 strain). All detected strains are known to have pathogenic potential, particularly in immunocompromised individuals. Specifically, Escherichia coli can cause a range of infections, Pseudomonas aeruginosa is a common cause of hospital-acquired infections, and Klebsiella pneumoniae is associated with severe infections, especially in healthcare settings.

Detection of pathogenic bacteria through PacBio sequencing

We analyzed the microbial flora of sputum samples using third-generation PacBio sequencing, examining the data at the phylum, genus, and species levels. A total of 1,458,608 raw CCS were generated from 45 samples. Data filtering, length filtering, and chimera removal yielded 1,426,463 Effective CCS. Each sample had an average sequence length of 1,462 bp, with 97.85% of Effective-CCS, indicating high sequencing quality and enhancing the credibility of subsequent analyses. Using Usearch software to cluster reads at a 97.0% similarity threshold, we identified 13,168 OTUs across 45 specimens, reflecting the richness of microbial species. Variations in OTU counts between samples further highlight the diversity of microbial communities.

The abundance of species was visually displayed using a heatmap, followed by cluster analysis based on abundance similarity. The heatmap in Fig. 1 displays the species abundance of the microbial community in sputum samples at the species level, based on clustering analysis.

Fig. 1.

Clustered heatmap of species abundance at the species level for 45 samples. The color gradient ranges from blue to red, indicating the abundance levels from low to high, respectively. The specific values (6, 4, 2, 0, −2, −4, −6) on the scale bar represent the corresponding abundance levels, with lower values indicating lower abundance and higher values indicating higher abundance. “Treat1” refers to the experimental group in our study. The sputum samples collected from the same patient at different time points are marked with distinct colors. Heatmaps group species by abundance into blocks, revealing similarities and differences in community composition through color gradients and clustering patterns. In the clustering results, color represents species abundance. Vertical clustering indicates similarity in abundance across samples, while horizontal clustering shows the similarity in abundance of individual species. This emphasizes the community composition’s similarities

The Shannon diversity index dilution curve is used to assess the relationship between sequencing read count and alpha diversity in the samples. As shown in Fig. 2, the Shannon diversity index dilution curve rapidly rises initially and then stabilizes as the number of sequencing reads increases. This trend suggests that the abundance and diversity of species in the test group samples gradually stabilized with increasing sequencing data, confirming the adequacy of the selected data.

Fig. 2.

Shannon dilution curve

Figure 3 shows that bacterial species contributing over 1.0% of the total sequences are primarily from other genera (64.90%), with Lautropia mirabilis (5.06%), Rothia mucilaginosa (4.96%), Prevotella melaninogenica (4.40%), Granulicatella adiacens (3.38%), Streptococcus salivarius (3.26%), Streptococcus infantis (3.20%), Neisseria mucosa (2.99%), Stenotrophomonas maltophilia (2.69%), Lactiplantibacillus plantarum (2.68%), and Neisseria flavescens (2.48%) comprising the remaining 35.10%, collectively making up 100% of the total sequences. Notably, Lautropia mirabilis and Rothia mucilaginosa are the most abundant species, together comprising 10.02% of the total microbial community. Rothia mucilaginosa, a common resident of the human nasopharynx, oropharynx, and upper respiratory tract, can be isolated from nasopharyngeal secretions and bronchial aspirates. However, Rothia mucilaginosa has historically been relatively understudied, and its identification was not included in several older identification systems, such as Vitek, Vitek 2, and API Staph 32 version 2.0. Only the ATB ID 32 Staph and BBL Crystal (Becton Dickinson) systems can identify Rothia mucilaginosa. These limitations highlighted the challenges in its clinical detection. With the widespread adoption of MALDI-TOF MS, the identification of Rothia species has become more accurate and reliable, consistent with the findings of this study. The PacBio sequencing results highlight the remarkable bacterial diversity, reflecting the complex characteristics of the respiratory microenvironment in HIV co-infected patients. Consequently, there is a critical need for advanced diagnostic tools to accurately characterize and manage the microbial landscape in these complex clinical contexts.

Fig. 3.

Species-level composition of the sputum microbiota in 45 sample via PacBio sequencing

To ensure a comprehensive yet focused analysis, we selected the top ten most abundant bacterial species from each individual patient sample based on their CCS (Circular Consensus Sequence) counts. This approach allowed us to capture the most significant contributors to the microbial community in each sample while maintaining a manageable scope for detailed analysis. By aggregating these top species across all samples, we identified a total of 65 unique bacterial species, which accounted for 329 strains. These 65 species were chosen because they represented the most abundant bacteria in each patient sample, ensuring that our analysis was representative of the dominant microbial populations. The detailed species and relative proportions of microorganisms from the first ten CCS counts per sample are provided in Supplementary Table S1 (The sputum samples collected from the same patient at different time points are marked with distinct colors). The top 10 microorganisms, as shown in Fig. 4, were: Rothia mucilaginos (24 strains), Granulicatella adiacens (21 strains), Streptococcus infantis (18 strains), Veillonella parvula (18 strains), Streptococcus parasanguinis (16 strains), Streptococcus oralis (16 strains), Prevotella melaninogenica (14 strains), Streptococcus salivarius (13 strains), Haemophilus parainfluenzae (11 strains), and Streptococcus mitis (11 strains). Among the 329 strains, 199 (60.49%) were Gram-positive bacteria, primarily Rothia mucilaginosa (24 strains) and Granulicatella adiacens (21 strains). The remaining 120 strains (39.51%) were Gram-negative bacteria, mainly Prevotella melaninogenica (14 strains) and Streptococcus parasanguinis (11 strains). Detailed information can be found in Supplementary Table S2.

Fig. 4.

The top 10 microorganisms based on the total sum of the top 10 CCS counts for each sample

In this study, 65 bacterial species were identified in the samples using PacBio 16 S rRNA full-length sequencing. Notably, Rothia mucilaginosa and Granulicatella adiacens, representative Gram-positive bacteria, comprised a significant proportion, likely due to their ability to colonize the respiratory tract and adapt to the environment. Rothia mucilaginosa is a gram-positive coccobacillus of the micrococcaceae family, commonly found in human oral and upper respiratory tract microbiota. Despite its low virulence, Rothia mucilaginosa is increasingly recognized as an opportunistic pathogen, primarily affecting immunocompromised hosts [34]. Rothia mucilaginosa should be considered as a potential cause of pneumonia in both immunodeficient and immunocompetent individuals, highlighting the importance of early diagnosis and timely antibiotic treatment [35]. Furthermore, Prevotella melaninogenica, a dominant Gram-negative species, exhibits notable abundance, likely linked to increased inflammatory responses, disruption of nutrient metabolism, and imbalances in immune regulation. As a resident of the human oral cavity, Prevotella melaninogenica primarily contributes to oral health issues [36, 37].

Haemophilus parainfluenzae occupies a secondary role in the Gram-negative bacterial community and, as a conditional pathogen, can become the primary cause of respiratory tract infections in immunocompromised individuals [38]. These findings suggest that the proliferation of specific bacterial species may be closely linked to the disease process in patients with HIV-related lung co-infection.

Comparison of positivity rates between PacBio sequencing and traditional culture methods

PacBio sequencing identified 329 microorganisms (based on the top ten CCS counts per sample) in 45 sputum specimens, while traditional culture detected only 6 microorganisms. The number of pathogenic strains identified by PacBio sequencing significantly surpassed those detected by traditional culture.

PacBio sequencing identified microorganisms in all 45 sputum specimens, yielding a 100% detection rate. However, it is important to note that the presence of a pathogen does not necessarily imply causation, especially in immunocompromised patients where the microbiota can be highly complex and dynamic. Some of the detected microorganisms may represent transient bacteria or normal flora rather than true pathogens. Of the 45 samples, 4 were positive for pathogens using traditional culture, yielding a positive rate of 8.89%. The positive rate of PacBio sequencing (100%) was significantly higher than that of traditional culture (8.89%). These results demonstrate that PacBio sequencing provides a higher detection rate and broader microbial coverage. The results suggest that PacBio sequencing is highly sensitive and accurate, offering a comprehensive view of the microbial composition in the samples. The high positive rate reflects PacBio sequencing’s high-throughput, deep sequencing capabilities and its unique advantage in detecting low-abundance and difficult-to-cultivate microorganisms. In contrast, traditional culture had a lower positive rate of 8.89%. This result is expected, as traditional culture is limited by factors such as culture conditions, duration, and microbial growth characteristics, which can hinder the detection of all microorganisms. Specifically, traditional culture methods may fail to detect microorganisms with slow or harsh growth conditions, as well as Un-culture or dormant microorganisms and those in symbiotic states.

We compared the results of PacBio sequencing with those of traditional culture methods for consistency. Four cases were positive by both PacBio sequencing and traditional culture, indicating consistent identification of pathogens. Forty-one cases were positive by PacBio sequencing but negative by traditional culture. This comparison shows that PacBio sequencing is consistent with traditional culture methods and offers a high positive detection rate.

Validation of pathogenic bacteria detection in 7 samples via DdPCR

Seven sputum samples from seven patients (Sample 17, 24, 26, 29, 36, 37, 39) were selected for ddPCR validation based on results from PacBio sequencing and traditional culture. Samples were selected based on the criterion that five Sputum samples (Sample 17, Sample 24, Sample 26, Sample 29, Sample 37) tested positive for pathogens in traditional culture. Two Sputum samples (Sample 36, Sample 39) tested negative in traditional culture, but PacBio sequencing detected a high proportion of pathogenic microbial sequences.

A total of 64 strains from 13 different microorganisms were detected within the target range at the species level using ddPCR technology. Among these, 34 strains (53.13%) were Gram-negative bacteria, with Acinetobacter baumannii (7 strains, 10.94%) and Escherichia coli (6 strains, 9.38%) being the most prevalent. Additionally, 15 strains (34.38%) were Gram-positive bacteria, predominantly Streptococcus species (6 strains, 9.38%) and coagulase-negative staphylococci (CNS, 5 strains, 7.81%). Furthermore, 6 strains of Candida species were detected.

The Pilot AMR-1 Panel not only identified pathogens but also detected two antimicrobial resistance (AMR) genes: mecA and blaNDM/blaIMP. The AMR-1 panel detected a total of 9 AMR genes across the 7 samples tested. Gram-negative bacteria predominated in the test, comprising 53.13% of the total, with Acinetobacter baumannii and Escherichia coli being the most frequently detected. These bacteria, particularly Acinetobacter baumannii, are significant in clinical infections due to their high drug resistance and environmental adaptability, earning them the status of “superbugs” in hospital settings. Stenotrophomonas maltophilia and Escherichia coli are also broad-spectrum pathogens with notable drug resistance. Their high detection rates may indicate increased infection risk or antibiotic pressure in the sampled environment.

It is important to note that the detection of AMR genes was performed directly from genomic DNA extracted from sputum samples using ddPCR. This method allows for the identification of AMR genes but does not provide information on which specific bacterial species carry these genes. Therefore, we cannot confirm the exact bacterial species responsible for the detected AMR genes based on our current methodology. Additionally, phenotypic AMR testing was not performed on cultured isolates from Samples 17, 26, and 29. The primary objective of our study was to evaluate the presence of AMR genes using molecular techniques, and phenotypic testing was beyond the scope of this study.

Despite their relatively low detection rate (34.38%), Gram-positive bacteria remain important within the microbial community. The presence of Streptococcus species and CNS suggests potential clinical relevance. Certain strains of Streptococcus can cause respiratory infections, endocarditis, and other diseases, while CNS are common pathogens in hospital-acquired infections. Additionally, 9 strains of Candida species were detected, indicating a potential risk of fungal infection. Candida species, particularly Candida albicans, are common in immunocompromised individuals, such as those with AIDS, aligning with existing literature on opportunistic infections [39]. Candida albicans is a yeast that resides in the human digestive and genitourinary tracts, typically maintaining a symbiotic relationship with the body. However, overgrowth in immunocompromised individuals can lead to infections, making detection and accurate clinical interpretation critical.

In AMR gene testing, the Pilot AMR-1 panel identified two key resistance genes: mecA, associated with methicillin resistance, and blaNDM/blaIMP, linked to New Delhi metallo-beta-lactamases (NDMs) and extended-spectrum beta-lactamases (ESBLs), which hydrolyze a wide range of beta-lactam antibiotics, contributing to treatment failure. It is important to note that the mecA gene can be carried by both Methicillin-resistant Staphylococcus aureus (MRSA) and Coagulase-negative Staphylococci (CNS). While our study detected the presence of the mecA gene, further species-specific testing was not performed to confirm the exact bacterial source. The identification of 9 AMR genes across 5 of the 7 samples tested underscores the growing challenge of multi-drug resistance. This highlighting the need for ongoing surveillance and the development of more effective antimicrobial therapies in clinical practice.

Comparison of traditional culture techniques, PacBio sequencing and DdPCR

Pathogens detected in seven sputum specimens using traditional culture, PacBio sequencing, and ddPCR are summarized in Supplementary Table S3. To provide a more comprehensive view, Supplementary Table S4 includes all detected species for each sample. Among the seven sputum specimens, the traditional culture identified three bacterial species, comprising six Gram-negative strains. These included Escherichia coli (50.00%, 3/6), Pseudomonas aeruginosa (33.33%, 2/6), and Klebsiella pneumoniae (16.67%, 1/6). All detected bacteria were Gram-negative, consistent with numerous studies on the pathogenesis of respiratory tract infections. Gram-negative bacteria, with their unique cell wall structure, often exhibit high drug resistance and viability, enabling them to easily colonize and cause infections in environments such as the human respiratory tract. Consequently, the high detection rate of Gram-negative bacteria in sputum specimens suggests their prevalence in respiratory tract infections. Of the six detected bacteria, Escherichia coli was the most prevalent (50.00%), suggesting it may be a common causative agent in these sputum specimens. Escherichia coli, a component of the normal intestinal flora, can become an opportunistic pathogen and cause respiratory infections under specific conditions, such as immunocompromise or antibiotic use. Additionally, Klebsiella pneumoniae and Pseudomonas aeruginosa are common respiratory pathogens, and their detection highlights the complexity of respiratory infection etiology.

PacBio sequencing identified a total of 77 strains from 51 reference microorganisms, Gram-negative bacteria dominated the results, accounting for 59.74% (46 strains) of the total. Among these, the dominant strains were Escherichia coli (5 strains, 6.49%), and Stenotrophomonas maltophilia (5 strains, 6.49%). This finding is consistent with the characteristics of microbial communities in respiratory and environmental samples, where Gram-negative bacteria often predominate due to their unique physiological properties and environmental adaptations. Among the remaining 338.96% (30 strains) of Gram-positive bacteria, Granulicatella adiacens (3 strains, 3.90%) and Lactiplantibacillus plantarum (3 strains, 3.90%) were prominent. The detection of Escherichia coli, Stenotrophomonas maltophilia, Pseudomonas aeruginosa, and Klebsiella pneumoniae among the Gram-negative bacteria highlights their significance in specific environmental and pathological contexts.

Among the five sputum samples with positive culture results, both third-generation PacBio sequencing and ddPCR yielded consistent positive results, with the primary pathogenic microorganisms identified remaining largely unchanged. This confirms the presence of multiple pathogenic microbial infections. In comparison to the conventional sputum culture method, which identified three species, ddPCR detected an additional 12 pathogenic microorganisms. Third-generation PacBio sequencing identified an even broader range, detecting 47 distinct species. These findings underscore the superior sensitivity of both PacBio sequencing and ddPCR technologies, enabling the detection of a much wider array of pathogenic microorganisms than conventional culture methods.

Among the two sputum samples with negative culture results, both third-generation PacBio sequencing and ddPCR detected multiple pathogenic microorganisms. Sample 36 (culture negative): Both PacBio sequencing and ddPCR detected Streptococcus species, Klebsiella pneumoniae and Escherichia coli, along with other dominant species. Sample 39 (culture negative): Both methods identified Klebsiella pneumoniae as the primary pathogen, which was later confirmed through targeted sequencing of the sputum. These findings demonstrate that third-generation PacBio sequencing and ddPCR not only offer accuracy comparable to traditional culture methods but also provide valuable supplementary information when culture results are negative. The results highlight the high sensitivity and broad detection capability of these techniques, which could play a crucial role in clinical diagnostics. By enabling faster and more accurate pathogen identification, these methods could facilitate more targeted treatment, reduce unnecessary antibiotic use, and help combat antibiotic resistance. Moreover, these techniques may assist in discovering new pathogens or variants, contributing to better disease prevention, control, and public health safety.

Discrepancies between PacBio sequencing and ddPCR results can be attributed to differences in sensitivity, detection limits, and bioinformatics filtering criteria. For Sample 17, the PacBio data did not detect CNS and E. coli, while ddPCR showed CNS as positive. For Sample 26, PacBio sequencing did not detect CNS, while ddPCR showed CNS as positive. This discrepancy could be due to the higher sensitivity of ddPCR and stringent filtering criteria applied during the bioinformatics analysis of PacBio data. For Samples 24, 29, and 36, the mecA gene was detected by ddPCR, indicating the presence of methicillin-resistant Staphylococcus species. However, these species were not among the top 10 detected by PacBio sequencing. Supplementary Table S4 shows that Staphylococcus species were detected but were below the threshold for inclusion in the top 10 species list.

Comparison of diagnostic performance of PacBio sequencing and DdPCR

Comprehensive clinical judgment was made based on the patient’s clinical manifestations, laboratory tests, imaging, and traditional microbial culture results to confirm the presence of infection. A positive clinical diagnosis was established when infection was confirmed. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of PacBio sequencing and ddPCR were then evaluated. Among the 7 patients, both PacBio sequencing and ddPCR demonstrated 100% sensitivity, 0% specificity, 71.43% PPV, and 100% NPV. These results underscore the feasibility of these technologies for identifying pathogenic microorganisms. Table 2 compares clinical diagnostic concordance (positive and negative) between PacBio sequencing and ddPCR in sputum, while Table 3 presents their diagnostic performance metrics.

Table 2.

Comparison of positive and negative consistency between PacBio sequencing and DdPCR in sputum samples

		Clinical diagnosis					Clinical diagnosis
		Positive	Negative	Total			Positive	Negative	Total
PacBio sequencing	Positive	5	2	7	ddPCR	Positive	5	2	7
	Negative	0	0	0		Negative	0	0	0
	Total	5	2	7		Total	5	2	7

Table 3.

Diagnostic performance of PacBio sequencing and DdPCR in confirmed and undiagnosed pulmonary infection patients

Grouped	Sensitivity(%)	Specificity(%)	PPV(%)	NPV(%)
PacBio sequencing	100	0	71.43	100
ddPCR	100	0	71.43	100

Both third-generation PacBio sequencing and ddPCR technology demonstrated remarkable 100% sensitivity, indicating their exceptional ability to detect pathogenic microorganisms accurately in patients with confirmed infections. This highlights their high sensitivity and potential as reliable diagnostic tools, reducing the risk of missed diagnoses. However, it is important to note that both technologies had a specificity of 0%, meaning they struggled to accurately identify healthy individuals without infections. Specificity refers to the ability of a test to correctly exclude those without the disease, and the observed 0% specificity may be attributed to factors such as limitations in sample selection or experimental design. In clinical settings, low specificity can lead to false positives, where pathogens are incorrectly identified in healthy individuals.

Furthermore, the study’s findings, based on a small sample size and sputum specimens only, may not fully reflect the broader performance of these technologies in larger or more varied sample types. The PPV of both technologies was 71.43%, meaning that 71.43% of positive cases accurately represented true infections, demonstrating their reliability in identifying patients with infections. Additionally, both technologies exhibited a perfect NPV of 100%, confirming their accuracy in ruling out infections in negative cases.

It is crucial to acknowledge that the comparison of diagnostic performance between PacBio sequencing and ddPCR in this study is limited by the absence of a definitive gold standard for the presence or absence of specific pathogens in the patient samples. The comprehensive clinical judgment used in this study, while valuable, does not provide a clear and unambiguous reference for the “true” presence of bacteria. Therefore, the results of this comparison should be interpreted with caution and considered as preliminary findings. Future studies should aim to incorporate more definitive gold standards, such as culture-independent molecular methods or additional confirmatory tests, to provide a more robust basis for evaluating the diagnostic performance of these technologies.

Tables 2 and 3 provide a comparative analysis of PacBio sequencing and ddPCR, focusing on their clinical diagnostic concordance and performance. Despite differences in specificity, both technologies reached similar conclusions in most cases, with high sensitivity and NPV reinforcing their complementary strengths. Although the relatively small sample size limits generalizability, these findings offer valuable insights into the potential of these diagnostic tools, laying the groundwork for further research in larger, more diverse patient populations.

Discussion

Opportunistic infections are caused by pathogens (bacteria, viruses, or fungi) that typically do not lead to disease in individuals with a healthy immune system. However, HIV-positive individuals are particularly vulnerable to these infections due to their compromised immune function. Pulmonary infections, in particular, are among the most common and serious opportunistic infections in AIDS patients, contributing significantly to morbidity, mortality, and hospitalization. Traditional methods of pathogen diagnosis, such as microbial cultures, are often time-consuming, have low positivity rates, are limited by difficulties in obtaining sufficient pathogenic evidence. Culturing microorganisms can be particularly challenging in clinical settings, especially for HIV-coinfected patients with lung infections. Therefore, early and accurate diagnosis of pathogenic microorganisms in these patients is essential for improving clinical outcomes, enabling more precise diagnoses, and facilitating tailored treatment strategies [40].

Recent advancements in diagnostic techniques have significantly enhanced our ability to detect pathogens in immunocompromised hosts. For instance, a retrospective study evaluated the effectiveness of next-generation sequencing (NGS)-assisted pathogen detection in immunocompromised hosts with pulmonary infections, highlighting the potential of these techniques to identify pathogens missed by traditional culture methods [41]. Another study focused on the use of metagenomic NGS for pathogenic diagnosis and antibiotic management in severe community-acquired pneumonia in immunocompromised adults, demonstrating the clinical utility of these approaches [42]. Additionally, a comparison of blood pathogen detection among droplet digital PCR (ddPCR), metagenomic NGS, and blood culture in critically ill patients with suspected bloodstream infections further underscored the advantages of these molecular techniques over traditional methods [43].

With the continuous advancement of modern sequencing technologies, third-generation sequencing has shown considerable potential in microbiota research, thanks to its long read lengths and high accuracy. This study utilized third-generation PacBio sequencing to analyze the microbiota of sputum specimens collected from AIDS patients with lung infections, aiming to investigate the microbial composition and variations lung infections within this patient population. A total of 45 sputum specimens from 37 AIDS patients with lung infections were included in the study.

Analysis of microbial strains detected by PacBio sequencing revealed that Rothia mucilaginosa and Granulicatella adiacens were the predominant species in the samples. The study highlighted that Granulicatella adiacens is a conditionally pathogenic species, which can become pathogenic when the immune system is weakened. It has been associated with infections such as bacteremia, infective endocarditis, and otitis media [44, 45].

In our study of initial culture methods for immunocompromised patients, we carefully selected culture media, excluding MacConkey agar and fungal culture agar for specific reasons. We chose blood agar and chocolate agar based on their unique advantages. Blood agar, a nutrient-rich and versatile medium, supports the growth of a wide range of bacterial pathogens, including both Gram-positive and Gram-negative species. It is essential for isolating common respiratory pathogens such as Staphylococcus aureus and Streptococcus pneumoniae, and its ability to detect hemolytic reactions aids in pathogen identification. Chocolate agar, designed for fastidious organisms, is essential for isolating Haemophilus spp. and Neisseria spp., which are commonly found in respiratory samples and difficult to grow on other media. Although blood agar and chocolate agar are widely used and effective for culturing a broad range of pathogens, our study has limitations in detecting certain microorganisms. The conventional culture methods employed, including selective colony picking and relatively short incubation periods, may have led to the underestimation of some pathogens. For instance, the detection of fungi could have been enhanced by using specialized media such as Sabouraud dextrose agar. Additionally, the use of other selective media like MacConkey agar for Gram-negative bacteria and extended incubation periods could improve the detection rates of less common or slow-growing pathogens. Future studies should consider incorporating these additional selective media and longer incubation periods to provide a more comprehensive microbiological profile, especially for high-risk patient groups such as BMT or AIDS patients.

This study statistically analyzed sputum specimens, focusing on the number and diversity of microbial strains, and compared the strain detection capabilities of PacBio sequencing with traditional culture methods. The results showed that while the traditional culture method was able to detect common clinically relevant bacteria, such as Escherichia coli, Pseudomonas aeruginosa, and Klebsiella pneumoniae, its detection scope was limited. In contrast, the PacBio sequencing not only identified these common strains but also uncovered a broader range of other bacterial species, providing a more comprehensive and in-depth view of the microbial community. PacBio sequencing demonstrated a higher positive rate and a wider detection range for microbial pathogens. A comparison of the results from PacBio sequencing and traditional culture methods revealed strong consistency, with PacBio sequencing providing a broader and more detailed microbial profile.

Classical microbial culture methods, while effective in detecting common bacterial pathogens, have significant limitations, especially in the context of immunocompromised patients. These traditional techniques often prioritize the detection of fast-growing enteric bacteria, such as Escherichia coli, Pseudomonas aeruginosa, and Klebsiella pneumoniae. This bias can lead to an underestimation of other clinically significant pathogens, particularly slow-growing or fastidious bacteria. For instance, acid-fast bacteria like Mycobacterium tuberculosis, which are critical in managing opportunistic infections in AIDS patients, are often not detected by these methods. In our study, the results are consistent with these limitations. Traditional culture methods successfully identified common clinically relevant bacteria but failed to detect acid-fast bacteria or other slow-growing pathogens. This highlights the need for more comprehensive and sensitive diagnostic approaches to ensure accurate pathogen detection in this vulnerable patient population. Our study demonstrates the effectiveness of traditional culture methods in detecting bacterial pathogens in sputum samples from immunocompromised patients. However, we acknowledge that the incubation period of 1–3 days may not be sufficient for the detection of all potential pathogens, particularly slow-growing bacteria or those requiring specific enrichment conditions. This limitation could potentially affect the comparability of traditional culture results with PacBio sequencing results. Future studies may consider extended incubation periods or the use of enrichment broths to further enhance the detection of some slow-growing bacteria or fastidious organisms.

PacBio sequencing enables the detailed characterization of the microbial flora in patient sputum samples, providing insights into the microbial composition and relative abundance of species. However, it does not allow for the quantification of the absolute number of bacteria, meaning it cannot determine the exact concentration of specific bacteria present in the sample. On the other hand, ddPCR has shown promise as a diagnostic tool for bloodstream infections (BSIs) [46–48]. However, there have been no reports on the application of ddPCR for identifying pathogenic microorganisms in the lungs of HIV-coinfected individuals, leaving its role in lung infection diagnostics unclear. In this study, seven sputum samples from AIDS patients with lung infections were selected, and ddPCR was used to detect 18 specific bacterial species at both genus and species levels. Additionally, ddPCR was used to identify five drug-resistance genes, providing valuable quantitative data on microbial presence and antimicrobial resistance.

A comparison of traditional culture method, PacBio sequencing, and ddPCR for detecting pathogenic microorganisms in sputum DNA revealed that both PacBio sequencing and ddPCR exhibited superior sensitivity and a broader detection range than traditional culture. Both technologies demonstrated high detection rates in samples that were both positive and negative by traditional culture, with consistent identification of the primary pathogenic microorganisms. Specifically, PacBio sequencing detected 77 microbial strains, ddPCR detected 64 strains, and traditional culture identified only 6 strains. Notably, ddPCR also fungal species and drug-resistant genes, underscoring its expanded diagnostic capabilities and ability to identify a wider array of pathogens, including those with antimicrobial resistance. K. pneumoniae and P. aeruginosa were not detected in sample 36 through PacBio sequencing. K. pneumoniae and P. aeruginosa were detected in sample 39 through PacBio sequencing while traditional culture result was negative. We acknowledge that K. pneumoniae and P. aeruginosa are generally considered to be easily cultivable species. However, there are several potential reasons why these pathogens might not have been detected in traditional culture, despite being identified through PacBio sequencing. First, sputum samples can be heterogeneous, with different regions of the sample containing varying microbial compositions. It is possible that the portion of the sample used for traditional culture did not contain sufficient quantities of K. pneumoniae or P. aeruginosa to be detected, while the sequencing analysis, which processes a larger volume of the sample, was able to identify these pathogens. Second, traditional culture methods have inherent detection limits. If the abundance of K. pneumoniae or P. aeruginosa was very low in the sample, they might have been below the detection threshold of the culture method but still detectable through sequencing. Third, if the patients had recently received antibiotic treatment, the pathogens might have been present in a viable but non-culturable (VBNC) state. These bacteria can still be detected through sequencing but may not grow in traditional culture due to residual antibiotic effects.

Although PacBio sequencing offers significant advantages in terms of microbial identification accuracy and the ability to directly analyze genetic material, the practical application of this technology in clinical settings requires careful consideration of the balance between turnaround time (TAT) and running costs. In our study, the TAT for PacBio sequencing was approximately 1 month due to the involvement of multiple steps, including sample collection, transportation, pre-sequencing processing, and data analysis. However, it is important to note that many professional sequencing companies can deliver sequencing results within 24–72 h for routine projects. The extended TAT in our study was primarily due to the coordination and communication required across various stages of the project. In comparison to conventional culture methods, which often take days or even weeks to provide results, PacBio sequencing has the potential to significantly reduce TAT under optimized conditions. However, the running costs associated with PacBio sequencing, including sample preparation, sequencing, and bioinformatics analysis, can be relatively high. Therefore, to make PacBio sequencing practical for clinical microbiological identification, future studies should focus on optimizing workflows to minimize TAT while also exploring cost-effective solutions to reduce running costs. This balance is essential to ensure that the benefits of rapid and accurate identification can be realized without compromising patient care due to delays in diagnosis.

In the diagnostic performance study, both PacBio sequencing and ddPCR demonstrated a sensitivity of 100% in detecting pathogenic microorganisms seven patients. Although the specificity is 0%, the PPV and NPV were both high, suggesting that these technologies are highly effective in identifying true infections. Compared to traditional culture methods, PacBio sequencing and ddPCR not only increased the detection rate of pathogens but also identified microbial infections in sputum samples that were negative by conventional culture, further highlighting the enhanced diagnostic potential of these methods.

In this study, we aimed to compare the performance of different diagnostic methods for identifying microbial pathogens. While traditional culture methods are often considered the gold standard in clinical microbiology, they have limitations such as long turnaround times and the inability to detect certain fastidious or slow-growing microorganisms. Given these limitations, we did not designate a single gold standard method. Instead, we used a multi-method approach, comparing traditional culture methods, ddPCR, and PacBio sequencing. This approach allowed us to assess the strengths and weaknesses of each method comprehensively.

Detecting pathogens in opportunistic infections in AIDS patients is a highly challenging problem. Even with the use of traditional culture methods and advanced DNA-based methods, such as those employed in this study, determining whether a detected microorganism is the causative agent of an infection requires careful consideration. The presence of a pathogen does not necessarily imply causation, especially in immunocompromised individuals where the microbiota can be highly complex and dynamic.

In our study, we detected a variety of microorganisms using traditional culture methods, PacBio sequencing and ddPCR. While these methods provide valuable information on the presence of potential pathogens, they do not definitively establish causation. The determination of causative agents in opportunistic infections often requires additional clinical context, such as the patient’s symptoms, immune status, and response to treatment. Moreover, the presence of multiple potential pathogens can further complicate the interpretation of results.

To address this challenge, clinicians and researchers must integrate multiple lines of evidence, including clinical symptoms, laboratory findings, and microbiological data, to make informed decisions about the causative agents of infections. In the context of AIDS patients, the immune status and the specific opportunistic pathogens commonly associated with HIV/AIDS must be considered. For instance, certain pathogens like Pneumocystis jirovecii and Mycobacterium tuberculosis are well-known causes of opportunistic infections in AIDS patients, but their detection alone does not confirm active infection.

Conclusion

This study employed third-generation PacBio sequencing to analyze sputum samples from HIV-coinfected patients with lung infections, uncovering distinct microbiota characteristics. The accuracy and reliability of PacBio sequencing were validated through comparison with traditional culture methods, providing new insights for the clinical diagnosis and treatment of HIV-associated lung infections. The study also emphasized the complementary role of ddPCR in the rapid detection of common pathogens and antibiotic-resistant genes, which enhances the ability of PacBio sequencing to identify pathogens undetectable by conventional methods. Together, these technologies offer a more comprehensive and accurate approach to pathogen detection, with anticipated advancements expected to enhance precision medicine and infection control efforts. The primary objective of our study was to conduct a methodological comparison between different diagnostic techniques for detecting pathogens in immunocompromised patients. While comparing our results with those from healthy individuals or patients after treatment could provide additional insights, such comparisons were beyond the scope of this study. Future research should consider incorporating these comparisons to better understand the clinical implications and to distinguish between commensal bacteria and causative agents. Due to the retrospective nature of this study, complete information on CD4⁺ T-cell counts, inflammatory markers, and other immune-activation indices was not available for all participants. Although our study provides valuable insights into the microbial composition and potential pathogens in AIDS patients, it is essential to recognize the limitations of our methods and the need for careful clinical evaluation to determine causative agents of opportunistic infections.

Abbreviations

ddPCR

: droplet digital PCR

OIs

opportunistic infections

SMRT

Single-molecule real-time

CCS

circular consensus sequencing

mNGS

metagenomic next-generation sequencing

CNS

coagulase-negative staphylococci

AMR

antimicrobial resistance

MRSA

methicillin-resistant Staphylococcus aureus

NDMs

New Delhi metallo-beta-lactamases

ESBLs

extended-spectrum beta-lactamases

PPV

positive predictive value

NPV

negative predictive value

BSIs

bloodstream infections

TAT

turnaround time

Authors’ contributions

Conceptualization, X.Y. and J. L.; Methodology, X.Y. and J. L.; Data curation, Y. L., W. J. and T. Z.; Investigation, Y. L. and W. J.; Formal analysis, Y. L. and T. Z.; Validation, Y. L.; Resources, X.Y.; Writing - Original draft, Y. L.; Review and editing, W. J., X.Y. and J. L.; Supervision, X.Y. and J. L.; Project administration, X.Y. and J. L.; Funding acquisition, X.Y. and J. L.; All authors read and approved the final manuscript.

Funding

This work was supported by Natural fund project of Fujian Province, China (2024J011223); Fuzhou Health science and technology plan soft science research project (2022-S-wr4); Young and middle-aged talent research project of Fuzhou city (2022-S-rc5); Fujian Provincial Science and Technology Plan Project (2022Y4003); Fuzhou Science and Technology Project (2022S005); Fujian Provincial Major Health Research Project (2022ZD01001).

Data availability

The sequencing raw data is stored in the SRA database of NCBI, accession number PRJNA1196648.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the Medical Ethics Committee of Mengchao Hepatobiliary Hospital of Fujian Medical University with the approval number: 2020-088-01. Informed consent to participate was obtained from all participants in the study. This study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.