Tulane researchers develop faster, more accurate method to detect antibiotic-resistant bacteria.
Deadly infections caused by drug-resistant bacteria, specifically those immune to antibiotics, are becoming more common around the world, and they’re harder to treat than ever. Diseases like tuberculosis and staph infections are leading to longer hospital stays, higher costs, and in many cases, higher death rates. In 2021 alone, nearly half a million people developed tuberculosis that didn’t respond to standard medications. Even when people did get treatment, only about half recovered. This has pushed researchers to find new ways to spot these resistant bacteria faster, so doctors can treat them more effectively.
Scientists at Tulane University have come up with a new way to tackle the drug-resistant bacteria issue. They’ve created a computer-based method that looks at the bacteria’s genes to figure out which ones can fight off antibiotics. This model doesn’t need to know in advance which mutations cause resistance. Instead, it scans the entire DNA of the bacteria and finds patterns that match up with resistance. This makes it different from older methods that rely on already-known connections between certain genes and resistance. Older tools can make mistakes, linking the wrong genes to resistance, which can lead to incorrect diagnoses and the wrong medications.
The tool Tulane developed is based on machine learning, meaning the computer learns from examples. It studies bacteria with known resistance and spots the differences in their genetic makeup compared to bacteria that are still treatable with antibiotics. The researchers tested it on over 7,000 strains of tuberculosis and almost 4,000 samples of staph bacteria. The results were impressive. The model was just as accurate—if not more—than the World Health Organization’s current method. And it made far fewer mistakes when it came to flagging bacteria as resistant when they weren’t.
One of the problems with current testing methods is time. Some tests, like growing the bacteria in a lab, can take weeks. Others, like quick DNA tests, can miss rare forms of resistance. Tulane’s model solves both problems. It can spot rare mutations and work much faster, using the bacteria’s entire genetic blueprint to give a clearer picture.
A big part of the success came from the way the model was trained. It didn’t just guess based on a few examples. Instead, it looked at large groups of bacteria, comparing drug-resistant bacteria with those that weren’t immune. That way, it could find solid clues about which mutations really mattered. And because it doesn’t need human experts to guide it, this approach could be used in more places, including places that don’t have access to high-level labs.
Tests run in China using real patient samples showed that this new model worked better than the WHO’s current system. That means doctors could know sooner whether a patient needs a different kind of treatment, which is critical in stopping infections from getting worse or spreading to others.
This new tool also has potential outside hospitals. It could one day be used to track resistance in farming, where overuse of antibiotics in animals can lead to the spread of resistant bacteria. For now, though, the focus is on helping patients get the right treatment faster and cutting down on the use of medications that might not work.
One of the researchers summed it up well: the best way to stay ahead of these superbugs is to keep improving how we spot them. And this tool is one more way to do that.
Sources:
AI model enhances detection of antibiotic resistance in deadly bacteria
Join the conversation!