Biases, blind spots and signals in public One Health genome data for carbapenemase-producing Enterobacterales

Imagine the world as a giant, invisible web connecting people, animals, the food we eat, and the water we drink. This is the "One Health" concept: the idea that our health is deeply tied to the health of animals and our environment. Now, imagine a super-villain hiding in this web: tiny bacteria that have learned to ignore our strongest antibiotics. These are called Carbapenemase-producing Enterobacterales (CPE). Think of them as the "boss monsters" of the bacterial world because they can shrug off the last-resort medicines doctors use when everything else fails.

To catch these villains, scientists use a powerful tool called Whole Genome Sequencing. It's like taking a microscopic photograph of the bacteria's entire instruction manual (its DNA) to see exactly how it fights back. These photos are uploaded to massive public libraries on the internet, like a giant digital museum where anyone can look at the evidence. But here's the catch: just because a photo is in the museum doesn't mean it represents the whole world. If the museum only has pictures of villains from one specific city, we might think that city is the only place they exist, even if they are hiding everywhere else. This is where the story of "bias" comes in—understanding what we see in the data versus what is actually happening in nature.

---

The Great Digital Detective Story: What's Hiding in the Bacterial Library?

Georgios Miliotis, a researcher from the University of Galway, decided to play detective with one of the world's biggest bacterial libraries: the NCBI Pathogen Detection Isolates Browser. He wanted to see if the library was telling the whole story about these super-bacteria across the "One Health" web, or if it was just showing us a skewed picture.

He zoomed in on 139,466 records of these super-bacteria. When he looked at the "source" tags—where the bacteria were originally found—he found a massive imbalance. It was like walking into a library expecting to see books from every corner of the globe, only to find that 84.72% of the shelves were stacked with books from hospitals and human clinics.

The other corners of the web were barely represented. The "wastewater" section had only 2.54% of the records. "Animals" had a tiny 1.30%, "food" had a microscopic 0.09%, and the rest of the environment was even smaller. The author suggests that this doesn't necessarily mean these bacteria are rare in animals or food; it just means that scientists and labs have been much more eager to sequence and upload the human hospital samples than the samples from a cow's gut or a river. The library is full of human stories, leaving the animal and environmental chapters largely unwritten.

The Hidden Patterns: Who Likes What?

Even with this lopsided library, Miliotis found some fascinating patterns that acted like clues. He noticed that certain types of bacteria seemed to have favorite "neighborhoods" based on the specific resistance gene they carried (the "carbapenemase family").

• The Animal Connection: When the bacteria were found in animals, they were most likely carrying the NDM gene. Specifically, the bacteria E. coli and Klebsiella pneumoniae were the main players here. The data showed that animal-associated NDM records were 2.33 times higher than you would expect if the bacteria were randomly distributed.
• The Wastewater Connection: In contrast, the wastewater samples were dominated by the KPC gene. The bacteria K. pneumoniae, Enterobacter cloacae, and Citrobacter freundii were the stars of this show. Wastewater-associated KPC records were 1.86 times higher than expected.

It's as if the NDM gene is the "animal lover" of the bacterial world, while the KPC gene is the "sewage dweller." The study found that these associations were statistically significant, meaning they weren't just random accidents, but the effect size was modest. This suggests that while there are preferences, these bacteria are still quite good at moving between different environments.

The "Blind Spots" and the Truth About Numbers

The paper is very careful to point out a major blind spot: The number of records in the database does not equal the number of bacteria in the real world.

If you see 1,000 photos of a specific type of bacteria in a database, it doesn't mean there are 1,000 of them in nature. It might just mean that a lab in China or the UK decided to sequence and upload 1,000 of them, while a lab in another country didn't upload any. The author found that the non-human records were heavily concentrated in just a few countries like China, the UK, the USA, and Germany.

The study explicitly rules out the idea that these public databases can tell us the "prevalence" (how common something is) of these bacteria. The author argues that we cannot use these record counts to say, "KPC is rare in food," because the database only had 130 food records. That low number might just mean no one looked, not that the bacteria aren't there.

The Takeaway

So, what's the verdict? The public genome data is a treasure trove that can help us spot patterns and generate new ideas about how these super-bacteria move between humans, animals, and the environment. It successfully highlighted that animal samples tend to carry NDM genes while wastewater samples lean toward KPC genes.

However, the author warns us not to get too excited about the raw numbers. The library is currently "human-heavy." To truly understand the One Health web, we need to fill in the missing pages. We need more sequencing of animals, food, and water, and we need better labels (metadata) so we know exactly where each sample came from. Until then, the database is a great map of what we have looked at, but it's not a complete map of where the bacteria actually live.