Metagenomics Explained: How Gene Sequencing Detects Hidden Infections

What metagenomic sequencing actually does

Standard infection tests ask one question at a time: is this a flu virus, is this E. coli, does this culture grow a known bacterium. Metagenomic next-generation sequencing, or mNGS, works the opposite way. It extracts every fragment of DNA and RNA present in a blood, spinal fluid or tissue sample and sequences all of it, then compares the result against reference databases to identify whatever organisms are there, according to a review of the technique published on the NCBI's PubMed Central platform.

Because the test does not require a prior hypothesis about the culprit, it can flag bacteria, viruses, fungi and parasites in a single run, including pathogens a clinician had not considered, the review states.

From the discovery of Covid-19 to a cancelled transplant

The clearest public demonstration of the technique came in late 2019, when Chinese virologists investigating unexplained pneumonia cases in Wuhan could not identify a cause with conventional tests. Switching to metagenomic sequencing showed the genetic material matched the coronavirus family, a finding reported by BBC Future that accelerated diagnostic and vaccine work on what became known as Covid-19.

Individual patient cases cited by BBC Future illustrate the stakes. At the University of California, San Francisco, a woman with meningitis and liver failure was due for a liver transplant when metagenomic testing identified hepatitis E virus instead. "We gave her an antiviral medication, she got better, and we spared her from undergoing the transplant," said Charles Chiu, a professor at UCSF quoted in the report. In a separate pilot at a London hospital, the same approach identified tuberculosis in a patient who had been presumed to have lung cancer, avoiding unnecessary surgery, according to the same report.

Faster, but not yet routine

Turnaround time has been a major barrier. Niranjan Nagarajan of Singapore's National University told BBC Future that metagenomic analysis used to take "several days to a week"; sequencing machines can now produce DNA results in one to two hours, with artificial intelligence tools speeding up the pattern-matching step.

The PubMed Central review gives more granular figures: conventional diagnostic workflows typically take 24 to 72 hours, while nanopore-based sequencing — a long-read technology — delivered same-day results in 86% of cases, with a median turnaround of 6.7 hours. Per-sample sequencing costs run from roughly $100 to $300 depending on the platform, the review says, though that figure excludes labour, computing and storage costs that add to the real price of running the test.

Where it outperforms standard tests — and where it still struggles

The same review compiled results from multiple clinical studies comparing mNGS with conventional diagnostics across several infection types.

These gains come with technical caveats the review flags as unresolved. In blood or spinal-fluid samples, over 90% of sequencing reads can come from the patient's own cells rather than a pathogen, masking weak microbial signals. In samples with very few microbes to begin with, contamination from lab reagents can account for up to 45% of the reads reported. The review also notes there is no agreed read-count threshold for calling a result positive, and that a single change to that threshold can flip a borderline sample from positive to negative.

Regulatory status compounds the uncertainty: in the United States, most mNGS tests run as laboratory-developed tests certified under CLIA rather than tests approved by the Food and Drug Administration, there is no dedicated insurance billing code for the procedure, and more than 23% of related insurance claims are denied, according to the review.

Using sequencing to catch outbreaks inside hospitals

A related application is tracking how an infection spreads once it is inside a hospital, rather than diagnosing a single patient. At the University of Nebraska Medical Center, clinical microbiology fellow Shawn Freed is using long-read whole-genome sequencing combined with machine learning to compare bacterial samples collected from patients and flag potential outbreaks early, according to a UNMC newsroom report.

Freed's team has sequenced 40 Gram-negative bacterial isolates and 15 vancomycin-resistant Enterococcus samples from Nebraska Medicine patients hospitalized for more than two days, then run phylogenetic analysis to see how closely related the strains are. The UNMC report notes the team chose long-read sequencing partly because it is cheaper and better at detecting plasmids — the mobile genetic elements that often carry antibiotic-resistance genes — than short-read alternatives. Healthcare-associated infection outbreaks cost an estimated $45 billion a year in the United States, the report states, through extended hospital stays and additional antibiotic treatment.

Early warning beyond the hospital bed

Nagarajan's laboratories in Singapore are already applying the same sequencing principle to sewage and air samples to look for emerging viruses before they cause recognised outbreaks, according to BBC Future's reporting, a use case researchers describe as a possible early-warning system for future epidemics.

Across both the diagnostic and surveillance uses, researchers interviewed describe metagenomic sequencing as a complementary tool rather than a wholesale replacement for existing microbiology, pending further standardisation of thresholds, regulatory clearance and cost reduction.