For the record, I would have gladly read the full article if it hadn't annoyed me so much with this attempt to make sure I read to the end, which, let's be fair, nobody is going to do unless they're a nerd with an insatiable curiosity and a manic compulsion to read every bit of text on the page; but then that's the audience this article targets anyway, so why make it annoying to read? Either you find the subject fascinating and you'll read the entire article anyway, or you don't and you won't get past the title.
I see that human publications strive for engagement, not citations, so they try to suspend the answer on the longest string that fits on the page. Pictures are nice though!
A big part of what makes me happy is that this seems quite niche to me, but it's getting a ton of attention which suggests people care about this more than I realized. That feels good! This is a huge deal for biodiversity in a huge swath of the Pacific Ocean.
Gehman's colleague Melanie Prentice, also with the Hakai Institute and UBC, had begun reviewing the sequencing analysis. Scrolling through data on her laptop, an unmistakable pattern jumped out. Of the hundreds of microorganisms in the samples, enormous quantities of bacteria from the genus Vibrio seemed present in sick stars. The pattern was so strong, Prentice tells me, that she assumed she'd bungled the analysis.
[0] Trying to make sense of the hype around some of these: https://fortune.com/2025/07/06/deepmind-isomorphic-labs-cure...
Yes, we could get some broad strokes of what's present by using analytical tools, and we do. A lot of our researchers use R and Python to write simple tools to find patterns and perform simple aggregations.
We're in the process of discovering how AI can assist us, and we've had plenty of success in creating models to aid with processing and detecting things like mussel and kelp beds in aerial imagery (drone, plane, and satellite), but I suspect in this case the data can't actually tell the full story to begin with so what we're actually looking for is more like clues, not answers. When answers are present, we tend to be simply presenting findings like which organisms were present and, to some extent, their apparent population density. This is really easy to do without AI. Some models might find interesting patterns, but it's hard to tell which ones or why, or how we'd train that ourselves. It's an interesting question.
I work on the genomics team so it's something I should be actively considering. The work is certainly more manual than it needs to be, so there's work to be done here for sure.
It's tricky to identify the caustive agent from sequencing data. Sometimes it sticks out like a sore thumb. Eg I recently worked on samples from a patient who died, it was very easy to identify the cause because it was in the same at massive quantities (reads).
But usually, it's trawling through a hundred hits across all domains of life deciding is it consensual, contamination, pathogenic or database error. It's usually inconclusive.
Oddly enough, llm bots seem to be quiet good at this task without even fine tuning.
(Flesh eating bacteria don’t literally eat the tissue, they release toxins that kill the tissue so there’s many different kinds that can cause necrotising fasciitis).
> Now, in August 2025, after another year-and-a-half of work confirming and reviewing their findings, Gehman and her team have published a peer-reviewed study in the journal Nature Ecology & Evolution identifying Vibrio pectenicida, a saltwater-loving bacterium that works its way into sea star fluids, as the likely “dominant pathogen responsible for sea star wasting disease.” Gehman’s team has tracked down the killer.
Restoring the kelp forest would sequester more than €1.5B worth of carbon every year. That’s because the kelp grows so rapidly and then is sucked down into the deep sea. Ideally the ecosystem itself would somehow receive money for its services—but short of that, some foundation on behalf of the ecosystem —and short of that, some organization that’d be prepared to use the money to reinvest in other ecosystems.
Some are able to easily overpower our immune systems.