Science1 publisher2 min readPublished
Helmholtz profiler Metax flags contamination by how reads spread across a genome
Helmholtz Centre for Infection Research scientists built Metax, a profiler that judges a microbial DNA match by how widely its reads spread across the genome. It targets the false positives current profilers can report, a problem sharpest in low-biomass samples, where host and contaminant DNA swamp the microbial signal.
The Scientist · Science desk
What happened
- Taxonomic profilers are still under development and out of common use, and current tools can report false positives and inaccurate abundance estimates.
- Reads can match the wrong organism because different microbes share highly similar DNA and some reference genomes contain contaminating sequences.
- Laboratory reagents can add traces of their own microbial DNA during sample preparation, a contamination source known as the kitome.
- Metax profiles bacteria, archaea, fungi, other eukaryotic microorganisms and viruses, and estimates how abundant each one is.
- The team tested it on simulated data and on real human microbiome, environmental, wastewater and clinical samples.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- capability A lab working up an infection of unknown kind could look for bacterial, fungal and viral candidates in one analysis, with abundance estimates attached.
- constraint McHardy positions Metax output as a complement to established diagnostic approaches, so a positive call is a lead for clinicians to check with existing tests.
- precedent Wastewater surveillance and microbiome biomarker studies rest on the same presence calls, so a lower false-positive rate would change which pathogens and patterns they report.
A sequencing read that matches a reference genome is weak evidence, taken alone, that the organism is in the sample. Zhi-Luo Deng, first author of the study and a scientist in Alice McHardy's group at HZI, described the extra check [12]. "An important piece of information is where these matching reads occur across the genome," Deng said [7]. "If a microorganism is truly present, we generally expect reads to be distributed across multiple regions of its genome. False-positive signals, in contrast, are often restricted to only one or a few local regions, with little or no support elsewhere." [6]
The design fits the database problems. A stretch of DNA shared by related species, or a contaminating fragment inside a reference assembly, can only draw reads to the place where that stretch sits [4][6]. A microbe that is really there leaves reads scattered across its genome [6]. Metax weighs that coverage pattern alongside the sequence match and uses it both to call which microbes are present and to estimate how much of each there is [2].
The coverage test asks whether a genome's DNA is present. Reagent contamination puts real microbial DNA into the sample [5]. By the same logic, a reagent microbe whose whole genome gets in would leave reads spread across its genome the way a genuine organism's do [1].
On performance, the public account rests on McHardy's summary. "Overall, Metax identified microorganisms more accurately and estimated their abundances more reliably than the other methods tested. This is particularly important for low-biomass samples, where microbial signals can be overwhelmed by large amounts of other genetic material, for example, from the host or contaminants," McHardy said [10]. "Metax can, in a sense, better find the microbial needle in the haystack." [15]
The thing this doesn't tell you is the size of the gain. The release does not name the competing profilers or give sensitivity, false-positive rates or abundance errors for any of the sample sets. "Overall" also leaves open whether the clinical samples improved as much as the simulated ones [9][10].
That split is where I'd judge the method. In acute, life-threatening infections, quickly characterizing the microbes in a patient sample can help guide diagnosis and treatment [13]. A win on simulated communities, where the right answer is built in, is a benchmark result. Fewer wrong calls on real patient material is the result routine diagnosis needs, and taxonomic profilers are not yet in common use [3].
What to watch
- Per-sample false-positive and abundance-error figures in the Cell paper for the clinical samples, set against named competing profilers.
- Whether Metax or its users add a step to catch reagent contaminants whose whole genomes enter the sample and so pass the coverage test.
- Independent groups benchmarking Metax on low-biomass clinical samples outside HZI.