mNGS Without Host Interference: Sepsis Study Results
Sepsis remains one of the leading causes of in-hospital mortality, and identifying the causative pathogen quickly enough to guide antibiotic therapy is a persistent clinical challenge. Blood cultures — still the gold standard — are negative in up to 50–80% of cases, especially after empirical antibiotic treatment has begun. Metagenomic next-generation sequencing (mNGS) can detect pathogens that cultures miss, but human DNA interference consumes over 99% of sequencing reads in unprocessed blood, burying microbial signals. This webinar presents results from a 45-patient sepsis study at Taipei Veterans General Hospital (TVGH) evaluating whether host depletion filtration can overcome this bottleneck without distorting the microbial composition.
Watch the full webinar:
Key Takeaways
- Host depletion filtration achieved 5.3× average enrichment of microbial reads compared to unfiltered samples, without altering pathogen composition
- mNGS with filtration detected pathogens at rates comparable to pre-antibiotic blood cultures, even in samples collected after empirical antibiotic treatment
- In culture-negative patients (50–80% of sepsis cases), mNGS identified additional pathogens including polymicrobial infections invisible to standard cultures
- Genomic DNA (gDNA)-based extraction with filtration produced significantly higher microbial reads per million than cell-free DNA (cfDNA)-based methods
- The complete workflow — filtration, extraction, sequencing, and bioinformatic analysis — targets a 24-hour turnaround from sample to report
The Host DNA Problem in Sepsis Diagnostics
The traditional diagnostic pathway for sepsis begins with blood cultures collected before antibiotic administration. In clinical practice, empirical antibiotics are started immediately — clinicians cannot wait for culture results when a patient is septic. A second culture is often drawn after treatment begins, but antibiotic exposure reduces culture positivity rates dramatically: in this study, positivity dropped from 50% (pre-antibiotic) to 24% (post-antibiotic).
mNGS offers a fundamentally different approach. Rather than growing organisms in culture, it sequences all DNA in the sample and identifies pathogens computationally. The method can detect bacteria, fungi, and viruses simultaneously, identify organisms that fail to grow in culture, and return results within hours rather than days. But the overwhelming presence of human genomic DNA — roughly 1,000× more abundant than microbial DNA per cell — means most sequencing capacity is wasted on host reads unless the human DNA background is removed first.
Study Design — 45 Sepsis Patients, Four Parallel Workflows
The TVGH team enrolled 45 sepsis patients (22 male, 23 female; mean age 71). Primary infection sites included urinary tract (most common), pulmonary, and intra-abdominal infections. For each patient, the team collected:
- Blood culture 1 (BC1): drawn before empirical antibiotic treatment
- Blood culture 2 (BC2): drawn after empirical antibiotics began
- 8 mL whole blood: split into two 4 mL aliquots — one processed through the Devin Fractionation Filter, one without
From each aliquot, two DNA extraction methods were performed: genomic DNA (gDNA) extraction from the cell pellet after centrifugation, and cell-free DNA (cfDNA) extraction from the plasma fraction. This produced four parallel mNGS datasets per patient — filtered gDNA, unfiltered gDNA, filtered cfDNA, and unfiltered cfDNA — enabling direct comparison of both the filtration step and the extraction method.
Filtration Results — 5.3× Microbial Enrichment Without Composition Bias
A critical concern with any host depletion method is whether it introduces taxonomic bias — selectively removing or enriching certain microbial species. The TVGH data showed that filtration did not alter the overall microbiome composition. Pathogen profiles between filtered and unfiltered samples maintained a near-1:1 correlation, confirming that the zwitterionic charge-based capture mechanism targets mammalian nucleated cells specifically, without trapping microorganisms.
The practical impact was substantial: filtered samples averaged 246 microbial reads compared to 33 in unfiltered samples — a 5.3-fold enrichment. This increase directly improves detection sensitivity for low-abundance pathogens that would otherwise fall below the noise floor.
Pathogen Detection Compared to Blood Culture
The study compared mNGS pathogen concordance against both BC1 (pre-antibiotic) and BC2 (post-antibiotic) culture results:
| Method | Concordance vs BC1 | Concordance vs BC2 |
|---|---|---|
| mNGS with filter | 33.9% | 81% |
| mNGS without filter | 47% | 63% |
The positive agreement rate of mNGS with filtration against all culture reports from each patient was 51%, compared to 35% without filtration. Notably, mNGS with filtration performed comparably to BC1 even though the mNGS sample was collected after antibiotic treatment — a significant practical advantage since clinicians rarely have the luxury of drawing samples before starting empirical therapy.
Beyond concordance, mNGS consistently detected additional organisms not reported by blood culture. Some patients showed polymicrobial infections involving six or seven species, including mixed bacterial-fungal or bacterial-viral co-infections that blood cultures would never reveal. Dr. Chia-Ming Chang emphasized that interpreting these multi-pathogen results requires clinical experience and correlation with patient symptoms to determine which organisms are clinically significant.
Genomic DNA vs Cell-Free DNA — Why the Extraction Method Matters
The four-way comparison revealed a critical finding about extraction methodology. The gDNA-based approach with filtration produced significantly higher total microbial reads per million and target reads per million (reads mapping to the culture-identified species) compared to unfiltered gDNA.
However, the cfDNA-based approach showed no enrichment with filtration. Both filtered and unfiltered cfDNA samples had similarly low microbial read counts — comparable to unfiltered gDNA.
Mengchu Wu, CEO of Micronbrane Medical, explained the mechanism: the Devin Filter removes intact human nucleated cells, eliminating the source of high-molecular-weight human gDNA. After filtration, the cell pellet from centrifugation is enriched for microbial cells and particles. But cell-free DNA fragments (~150 bp) circulating in plasma — both human and microbial — pass through the filter regardless, so cfDNA-based extraction sees no benefit from filtration.
This finding supports the gDNA-based PaRTI-Seq workflow: filter first to deplete host cells, then extract gDNA from the enriched pellet. The result is either more microbial reads at the same sequencing depth, or equivalent sensitivity at lower sequencing cost — both of which improve the economics of clinical mNGS adoption.
About the Speakers
Chia-Ming Chang, MD, is an attending physician in the Emergency Department at Taipei Veterans General Hospital (TVGH), Taiwan. His research focuses on sepsis diagnostics and the clinical application of metagenomic sequencing for pathogen identification in critically ill patients.
Mengchu Wu, PhD, is the co-founder, CEO, and chairwoman of Micronbrane Medical. She co-invented the zwitterionic host depletion filter technology and has co-authored more than 20 peer-reviewed publications in leading journals.
