Plasma microbial DNA sequencing catches hidden infections after liver transplantation

When a patient leaves the operating room after a liver transplant, the most dangerous weeks are often the ones immediately ahead. Bacterial infections strike roughly twenty to eighty percent of liver transplant recipients, and opportunistic pathogens are a leading cause of death during the first three years after surgery. Yet diagnosing infection in this window is notoriously difficult. The immunosuppressive drugs that keep the new organ alive also blunt the inflammatory responses that normally announce an infection, so fever and other clinical signs may be muted or absent. Postoperative bleeding and organ rejection can mimic infectious disease, while atypical opportunistic pathogens may be invisible to the tests clinicians routinely order. A research team in South Korea now reports that a sequencing technique designed to sweep the bloodstream for microbial DNA fragments may help close this diagnostic gap during the earliest, most vulnerable post-transplant period.

The study, published in Frontiers in Cellular and Infection Microbiology, was conducted at Severance Hospital, a large tertiary teaching hospital affiliated with Yonsei University College of Medicine in Seoul. Between January 2024 and June 2025, the investigators enrolled 216 adult liver transplant recipients in a prospective diagnostic cohort. Twenty-three patients were later excluded because a cell-free DNA sample could not be obtained on either of the two scheduled sampling days, leaving 193 participants in the final analysis. Each recipient underwent plasma microbial cell-free DNA metagenomic next-generation sequencing, abbreviated mcfDNA mNGS, on postoperative days 7 and 14. The technique works by isolating the small fragments of DNA that microbes shed into the blood, sequencing them without any prior assumption about which organism might be present, and then matching the resulting reads against reference genomes to identify bacteria, viruses, and fungi at the species level.

The laboratory workflow behind the assay is technically demanding. Whole blood was collected in specialized tubes that stabilize cell-free DNA, then centrifuged twice to separate the plasma. Cell-free DNA was extracted with a magnetic reagent kit and quantified with a high-sensitivity fluorescence assay before libraries were prepared and sequenced on an Illumina NovaSeq 6000 system. The resulting reads were first aligned to the human genome to strip away host DNA; the unmapped remainder was then fed into Kraken2, a fast taxonomic classifier running against a custom database built from NCBI bacterial, viral, and fungal genomes. To make results comparable across samples sequenced at different depths, the team converted raw read counts into a normalized metric called reads per ten million, or RPTM, and applied a uniform positivity threshold of RPTM at or above 5. Known environmental and skin organisms such as Cutibacterium acnes were masked, and organisms appearing in high abundance only within specific processing batches were manually excluded as extraction or library-preparation contaminants.

Because fungal detection is particularly prone to database-driven false discoveries, the researchers built an unusual cross-validation layer into their pipeline. Every fungal call from Kraken2 was re-examined with two auxiliary classifiers: Kaiju, which matches translated protein sequences, and EukDetect, which aligns fungal marker genes. Supporting reads were then retrieved from the host-depleted sequencing files and mapped directly against a custom reference panel of 416 fungal genomes, with the human genome included as a decoy. A candidate fungal detection was considered confirmed only when at least ten reads with high mapping quality were distributed across at least three distinct one-kilobase windows. This adjudication process confirmed 92.9 percent of the primary pipeline’s fungal detections, but it also revealed a serious blind spot: the primary classifier achieved high precision for the fungi it called yet captured only a small fraction of the alignment-confirmed fungal detections in the cohort, missing organisms including Candida species, Nakaseomyces glabrata, and Aspergillus niger. All of the missed detections were recovered by Kaiju, and a consensus across multiple classifiers confirmed every detection when all three pipelines agreed.

The headline results concern agreement with conventional microbiology. Among the 193 recipients, mNGS returned positive results in 14.5 percent of patients on day 7 and 21.8 percent on day 14. When the sequencing results for bacterial pathogens were compared directly with blood cultures, concordance appeared modest: 35.0 percent on day 7 and just 14.3 percent on day 14. But that comparison understates the assay’s real performance. When the researchers widened the reference standard to include cultures from other clinically relevant specimens collected within 14 days, such as sputum, peritoneal drain fluid, and urine, overall concordance rose to 80.0 percent on day 7 and 71.4 percent on day 14, a level the authors describe as clinically acceptable. The low direct agreement with blood cultures likely reflects both the higher analytical sensitivity of sequencing and the fact that most patients were already receiving broad-spectrum empirical antibiotics, which suppress culture growth but leave microbial DNA fragments circulating in the plasma.

The sequencing assay also uncovered pathogens that routine testing missed entirely. Bacteria were detected exclusively by mNGS in 59.3 percent of cases on day 7 and 74.2 percent on day 14. Fungal pathogens appeared in five and four cases on the two sampling days, with Aspergillus species identified only by sequencing on day 7. Viral detections were especially revealing: nine viral pathogens were found on day 7 and seventeen on day 14, including cytomegalovirus, herpes simplex virus, varicella-zoster virus, and roseolovirus. Cytomegalovirus detections were confirmed by quantitative PCR, but the other viruses are not routinely tested after liver transplant, and prophylactic strategies for them vary widely across countries. A retrospective chart review found that most patients with these unexpected viral detections had mucocutaneous manifestations suggestive of active infection, even though no diagnostic testing had been ordered and many had not received antiviral treatment. The authors suggest that these infections may be under-recognized under immunosuppression, and that sequencing could serve as an adjunctive diagnostic tool, while cautioning that some detected viruses might represent bystanders rather than true pathogens.

Discordant cases, where blood culture was positive but sequencing was negative, received detailed scrutiny. Thirteen such cases were identified. Two, or 15.4 percent, were classified as contamination or colonization, involving skin flora such as coagulase-negative staphylococci isolated from only one bottle of a blood culture pair or from a catheter specimen alone. Five, or 38.5 percent, were interpreted as weak positives, meaning the organism was present in the sequencing data but below the predefined RPTM threshold. The remaining six, or 46.2 percent, were classified as not detected; four involved Candida species, one Enterococcus faecium, and one Achromobacter species, and several of these were identified only from catheter blood cultures rather than peripheral blood. The Candida findings align with the pipeline’s documented limited sensitivity for fungi, reinforcing the authors’ argument that a single taxonomic classifier is insufficient for fungal surveillance and that negative sequencing results cannot reliably exclude fungal infection.

Sequencing positivity also tracked closely with clinical severity and survival. On day 7, patients with positive mNGS results had significantly higher plasma cell-free DNA concentrations, higher total bilirubin, higher C-reactive protein levels, and more impaired pre-transplant liver function than those with negative results. Their 28-day mortality was 26.9 percent versus zero percent among mNGS-negative patients, and their 90-day mortality was 30.8 percent versus 3.9 percent. On day 14, the 90-day mortality rate remained significantly higher among mNGS-positive patients. Among the 160 patients sampled at both time points, 15 of the 22 who were positive on day 7 remained positive on day 14, and the same organism persisted in 86.7 percent of those cases, indicating good reproducibility of the assay. Persistent positivity was associated with higher Sequential Organ Failure Assessment scores and elevated bilirubin and C-reactive protein levels, suggesting that the sequencing results mirror the recipient’s evolving clinical status rather than floating free of it.

The authors are careful to frame the technique as an adjunct rather than a replacement. Because mNGS is hypothesis-free and highly sensitive, it can detect DNA from dead or dying organisms, complicating interpretation and potentially prompting unnecessary antimicrobial use. Testing is also expensive, and the positivity rate in this unselected cohort was low, which the researchers argue supports selective deployment in situations where the expected diagnostic benefit is high, such as clinically ambiguous cases or patients already on broad empirical therapy in whom sequencing might facilitate early antibiotic de-escalation. Notably, most mNGS-positive patients in the study were already receiving active antibiotics against the detected pathogens, which limited the potential for the assay to trigger unnecessary new treatment. The study has limitations: sequencing was performed regardless of clinical suspicion, so positivity rates might be higher in suspected-infection settings; the comparison with blood cultures may have been skewed by prior antibiotic exposure; and the retrospective elements prevented assessment of whether sequencing-guided treatment improved outcomes. Cost-effectiveness, turnaround time, and antimicrobial resistance gene detection were not evaluated, and the authors suggest that differentiated thresholds, such as RPTM at or above 3 for viruses and at or above 8 for bacteria and fungi, might better balance sensitivity and specificity. Whether sequencing-guided management can improve clinical outcomes after liver transplantation remains the central question for future studies, but the evidence here indicates that plasma microbial cell-free DNA sequencing agrees substantially with conventional microbiology when clinically relevant specimens are considered, and that it reveals pathogens, particularly viruses and antibiotic-suppressed bacteria, that routine diagnostics would otherwise leave undetected during the fragile early weeks after surgery.

Subject of Research: Plasma microbial cell-free DNA metagenomic sequencing for early infection detection after liver transplantation

Article Title: Plasma microbial cell-free DNA metagenomic sequencing for early detection of infection after liver transplantation: a prospective cohort study

Article References: Lee, Y., Min, E.-K., Park, Y. J., Kim, D.-G., Joo, D. J., Kim, M. S., Shin, S., Lee, S.-T., Lee, J. G., & Jeong, S. J. (2026). Plasma microbial cell-free DNA metagenomic sequencing for early detection of infection after liver transplantation: a prospective cohort study. Frontiers in Cellular and Infection Microbiology, 16, Article 1930701. https://doi.org/10.3389/fcimb.2026.1930701

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1930701

Keywords: liver transplantation, metagenomic next-generation sequencing, microbial cell-free DNA, infection diagnosis, bloodstream infection, organ transplant recipients, opportunistic pathogens, diagnostic microbiology, viral detection, fungal surveillance, prospective cohort study, antibiotic de-escalation