TVGH Sepsis mNGS Study: PaRTI-Seq Detects 74.4% of Cases
Overview
Sepsis pathogen identification with traditional blood culture has well-documented limits — sensitivity drops sharply once empirical antibiotics begin, and the 24–48 hour turnaround forces clinicians to treat empirically far longer than they should. This research summary presents results from a clinical study at Taipei Veterans General Hospital (TVGH) evaluating the PaRTI-Seq™ workflow — metagenomic next-generation sequencing (mNGS) paired with the Devin™ Host Depletion Filter and Devin™ Microbial DNA Enrichment Kit — head-to-head against blood culture and mNGS without host depletion in 45 septic patients.
→ Download the Full Research Summary (PDF)
Key Takeaways
- PaRTI-Seq detected pathogens in 74.4% of septic patients — versus 50.0% for blood culture before antibiotics and 22.0% for blood culture drawn after antibiotics had begun
- Host depletion did not distort which organisms were detected: filtered and unfiltered samples showed R² = 0.96 microbial composition correlation
- The Devin filter increased microbial reads from 46 to 243 per million QC reads, with the strongest enrichment in low-concentration samples
- Sensitivity (vs BC2 reference): PaRTI-Seq 81.8% vs mNGS without filtration 63.6%
- Results available within 24 hours — enabling earlier pathogen-directed therapy compared to the 24–48h+ blood culture wait
Key Findings
- Data were analyzed from 50 patients, with complete results for 45.
- Positive rates: PaRTI-Seq detected pathogens in 74.4% of cases — the highest of the four methods tested. mNGS without filtration reached 51.1%; the first blood culture (BC1) reached 50.0%; the second blood culture (BC2), drawn after antibiotic exposure, dropped to 22.0%.
- Filtration effect: The Devin filter did not alter microbial composition (R² = 0.96 between filtered and unfiltered samples) but increased microbial reads from 46 to 243 per million QC reads. Enrichment was strongest in samples with lower microbial concentration.
- Sensitivity (using BC2 as reference): PaRTI-Seq reached 81.8% sensitivity versus 63.6% for mNGS without filtration.
- Workflow: gDNA-based mNGS with host depletion delivered higher diagnostic yields than traditional blood culture and cfDNA-based mNGS without host depletion, with results in 24 hours.
Study Design
Each enrolled patient had blood drawn at two timepoints: once on admission, before antibiotics (the BC1 draw), and again on consent after empirical antibiotic exposure (the BC2 draw). The first draw underwent conventional blood culture only. The second draw was divided into three portions: one for a second blood culture (BC2), and two for mNGS — one processed through the PaRTI-Seq workflow (including the Devin filter) and one without filtration. Both mNGS portions were spun by sequential centrifugation (400 × g for 15 min, then 16,000 × g for 15 min) to yield a gDNA pellet and a cfDNA supernatant. This design enabled a four-way head-to-head comparison: BC1, BC2, mNGS with PaRTI-Seq workflow, and mNGS without host depletion — across both gDNA and cfDNA fractions of the same sample.
The PaRTI-Seq pipeline added PaRTI-Seq Analysis on top of the wet-lab steps, applying spike-in normalization, system control subtraction, and clinical thresholds to convert raw reads into actionable pathogen calls. Spike-in controls (ZymoBIOMICS reference organisms at 10⁴ genome copies/mL) were added to every sample, including the no-template control, to quantify reagent background.
Sample-Level Results
Across eight illustrative cases included in the summary, species-specific gDNA reads consistently exceeded cfDNA reads — often by an order of magnitude. PaRTI-Seq recovered the clinically diagnosed pathogen in cases involving Escherichia coli, Klebsiella pneumoniae, Proteus mirabilis, MRSA, Pseudomonas aeruginosa, Achromobacter species, and a polymicrobial E. coli / Candida albicans infection. Notably, PaRTI-Seq detected pathogens in three samples where BC2 was negative — exactly the post-antibiotic scenario where culture loses sensitivity and clinicians most need a confirmatory result.
Why This Matters for Sepsis Care
The clinical penalty for missed pathogen identification in sepsis is well established: every additional hour of inappropriate empirical therapy increases mortality. Two findings here directly address that gap. First, PaRTI-Seq’s 74.4% positive rate against BC1’s 50.0% means roughly half the cases that culture missed were resolved by sequencing. Second, the gap widens after antibiotics: BC2 dropped to 22.0% while PaRTI-Seq held at 74.4%. This is the operational window where sequencing-based diagnostics have the most clinical leverage — when the culture has already failed and broad-spectrum coverage is being escalated.
The R² = 0.96 correlation between filtered and unfiltered microbial composition also matters. It rules out the most common objection to physical host depletion methods — that they bias the recovered community by selectively removing certain microbes. The Devin filter enriches the signal without distorting it, which is what host depletion needs to do for clinical diagnostic use. For broader context on this issue, see Host Depletion for Pathogen Identification in mNGS.
Frequently Asked Questions
Does host depletion change which organisms are detected?
No. The Devin filter enriched microbial reads without altering microbial community composition — filtered and unfiltered samples showed R² = 0.96 correlation. The filter increases the signal-to-noise ratio without introducing selection bias toward any particular organism.
How does PaRTI-Seq compare to blood culture once antibiotics have started?
Substantially better. Blood culture sensitivity drops sharply after antibiotic exposure: in this study, BC2 (drawn post-antibiotics) detected pathogens in only 22.0% of cases, versus 74.4% for PaRTI-Seq on the same draw. This post-antibiotic window is where sequencing-based diagnostics provide the greatest clinical leverage.
What is the turnaround time?
Results are available within 24 hours. Blood cultures typically require 24–48 hours for initial growth, and anaerobic organisms may take several additional days.
Does the sample fraction — gDNA versus cfDNA — affect results?
Yes. Across eight illustrative cases in the research summary, species-specific gDNA reads consistently exceeded cfDNA reads — often by an order of magnitude. The authors note that cfDNA may be more prone to noise, with a higher risk of false positives. gDNA-based mNGS with host depletion is the recommended approach.
Is this study peer-reviewed?
Yes. The study is published as Chen Y-C et al., Molecular Diagnosis & Therapy (2025), DOI 10.1007/s40291-025-00797-3, open access. The two-page research summary available for download distills the full paper’s methods and findings.
Download
Download the Full Research Summary (PDF)
The two-page summary includes the full experimental workflow diagram, the four-method positive-rate breakdown, the per-sample gDNA vs cfDNA read counts table, and the BC1/BC2 sensitivity comparison.
Peer-reviewed paper: Chen Y-C, Liao P-H, Chen Y-W, Yen DH-T, How C-K, Chang C-M. Optimization of Metagenomic Next-Generation Sequencing Workflow with a Novel Host Depletion Method for Enhanced Pathogen Detection. Molecular Diagnosis & Therapy (2025). DOI: 10.1007/s40291-025-00797-3 (open access).
Earlier source materials:
Products referenced in this research summary:
