By KIM BELLARD
Most of us can identify dogs from cats just by the sounds they make. We could probably even separate a dog’s bark from a wolf’s howl. If you are a nature lover, you might be able to identify different species of birds by their calls. If you are a cetologist, you might be able to separate the vocalizations whales make versus those dolphins make. Across the animal world, we’ve learned the different sounds that different species make, which has been useful in our survival.
But did you ever wonder if you can identify, say, e coli from other bacteria?
It turns out that you can, thanks to research at Delft University of Technology (TU Delft) in the Netherlands. Four years ago, they showed that bacteria made noise, which was, in itself, a startling finding (admit it: would you have ever guessed that?). They used a thin layer of graphene to create a graphene “drum” small enough to fit a single bacterium. Team member Cees Dekker observed: “What we saw was striking! When a single bacterium adheres to the surface of a graphene drum, it generates random oscillations with amplitudes as low as a few nanometers that we could detect. We could hear the sound of a single bacterium!”
The team used this finding to accomplish an important purpose: to find out if bacteria were resistant to specific antibiotics. If an antibiotic was applied and the sound continued; it hadn’t worked. If the sounds stopped, the bacteria had been killed.
The team wasted no time in creating a start-up – SoundCell – to commercialize the finding. It promised to identify the “right” antibiotic in one hour, rather than subjecting patients to rounds of different antibiotics in search of one the bacteria wasn’t resistant to.
The team isn’t resting on their laurels. Some of them got to wondering, huh, I wonder if different bacteria make different sounds. And, their latest research shows, not only do they but, through machine learning, those different species can be distinguished. Team lead Farbod Alijani says. “With this new study, we take a significant leap forward: we show that each bacterial species has its own nanomotion signature.”
Mind. Blown.
The researchers focused on three bacteria that are common in hospital settings: E. coli, S. aureus (which causes staph infections) and K. pneumoniae (which causes pneumonia). They tested two different machine learning models; one correctly classified the bacteria 87% of the time, and the other 88% of the time.
“By combining SoundCell’s existing antimicrobial testing prototype with this machine learning model, we can identify the bacterial infection and determine which drug is effective at the same time, based purely on the sound of a single bacterium,” says SoundCell CTO, Aleksandre Japaridze. Leo Smeets, physician microbiologist at RHMDC adds: “This approach eliminates the need for culturing, which normally takes days. And because the diagnostic steps are no longer performed sequentially, we can save even more time.”
“It’s a completely different way of interpreting the different species,” Dr. Japaridze says. “Not chemically or biologically, with markers and genes, but just purely on…mechanical behavior.”
Their paper concludes:
To sum up, our results show that combining the high sensitivity of graphene nanomotion sensors with ML enables fast, label-free AST and identification of bacteria. Since the trained models analyze nanomotion signals from individual cells, results can be obtained within 1-2 hours, eliminating the need for time-consuming culturing steps. With further development, this approach could establish nanomotion spectroscopy as a powerful platform for real-time diagnostics and for studying cellular biophysics and antimicrobial resistance.
They’ve been testing sensors in the lab, so one of the next steps is to show they can be used in actual hospital settings. They’re testing a prototype at two Dutch hospitals (RHMDC and Erasmus Medical Center). Professor Alijani believes: “This close partnership between scientists at TU Delft, a start-up and a hospital is quite unique. We have the entire knowledge chain working together.”
The potential impact is huge, with over 1 million deaths due to drug-resistant bacteria annually. “We have already shown that we can reduce antimicrobial susceptibility testing to one hour,” says Dr. Japaridze. “If we can combine that speed with species classification using the new machine learning model, we could create a globally unique device that dramatically accelerates diagnosis and treatment. And that would be highly valuable in the worldwide fight against antimicrobial resistance.”
———
I love the kind of curiosity that makes one wonder, hmm, do bacteria make noise? That’s not a question most people would ask themselves. I love the scientific expertise that figured out a way to actually detect that noise, at the level of a single bacterium. I love the realization that perhaps different bacteria make different noises, and the expertise to use machine learning to distinguish them. And, of course, I’m excited that all this might lead to practical applications that could save lives and avoid needless rounds of antibiotics.
Next thing you know, we might find out that bacteria not only make noise but use them to communicate. It wasn’t that long ago that we were arrogant enough to think that only humans communicate vocally, only to find that that many animal species use sound to communicate. Heck, we’ve even found that that plants “scream,” sending out messages we’re oblivious to.
It makes you wonder: what else are we missing?
I have this wild thought that our bodies are a cacophony, with all our cells and all of cells of our microbiota chiming in. When we’re healthy, perhaps they combine to create a finely tuned symphony, but when something is off it’s like an instrument in the symphony is badly tuned, off the beat, or missing. Perhaps if we listened the right way, we could use those sounds to more quickly and more accurately diagnose and treat the problem.
That’d be some 22nd century medicine.
So kudos to the scientists at TU Delft, good luck to the entrepreneurs at SoundCell, and to all you researchers in the world: keep asking these weird questions!
Kim is a former emarketing exec at a major Blues plan, editor of the late & lamented Tincture.io, and now regular THCB contributor
