Key Takeaways
Lyme disease, which is transmitted to humans by tick bites, is difficult to diagnose because it begins with symptoms such as headaches, pain, and fatigue and can develop into a long-term inflammatory disease that affects the joints, nerves, brain, and heart. The current gold standard for confirming Lyme disease is a two-step laboratory test that can take up to two weeks to produce results and often misses early-stage cases. A testing technology under development at UCLA uses artificial intelligence and a format similar to home COVID-19 tests, and new research shows that a single test is accurate enough to identify Lyme disease within 20 minutes.
Some unlucky people may contract Lyme disease while spending time in the wilderness. Lyme disease causes headaches, joint and muscle pain, flu-like symptoms, fatigue, and sometimes a rash. If left untreated, these symptoms can become debilitating and can escalate into paralysis, inflammation of the brain and heart, and memory, hearing, and vision problems that can last for years.
Lyme disease, which is expected to affect more than 600,000 people in the United States this year, is caused by a spiral-shaped bacterium that is transmitted to humans through the bite of an infected deer tick. The microbe triggers a complex immune response similar to that to other dangerous tick-borne bacteria, often making it difficult for doctors to start the appropriate antibiotic treatment.
To detect Lyme disease, the Centers for Disease Control and Prevention currently recommends a two-tiered testing schedule performed at a central laboratory. Results take one to two weeks to arrive. Antibiotic treatment doesn't cure all cases, but it can halt the progression of long-term disease in 80-90% of early-stage patients. However, current tests miss 7 in 10 early-stage cases, according to the Bay Area Lyme Foundation.
Members of the California NanoSystems Institute at the University of California, Los Angeles (UCLA) have developed a testing technology not dissimilar to home kits that detect COVID-19 infections. Results are interpreted within 20 minutes with a portable reader equipped with artificial intelligence. In a new study published in Nature Communications, the researchers used patient samples to show that the AI-enhanced one-shot test is as reliable as traditional Lyme disease testing methods that require two tests.
The method is rapid, portable and relatively low-cost, meaning it could potentially detect Lyme disease on the same day in the clinic where patients are treated.
“Many people don't realize they have Lyme disease until after it has been easily treatable,” said co-corresponding author Dino Di Carlo, the Armond Elena Hairpetian Professor of Engineering and Medicine at the Samueli School of Engineering at UCLA. “If we could measure it quickly in a way that was cost-effective and didn't burden the health care system or the patient, testing could be made more routine. If someone is out in the woods and has a tick bite or other symptoms, it would be wise to get a rapid test at home or at a local clinic, so they could potentially be treated sooner.”
A home test for COVID-19 involves applying a sample to a cartridge containing special paper for convenient comparison. The sample flows along the paper and passes antibodies that attach to proteins specific to the disease. The proteins are visualized by a second set of antibodies attached to tiny gold particles measuring a billionth of a meter. After a short wait, the results can be read by the human eye.
Lyme disease detection technology takes a different direction: A serum sample is dropped into a cartridge, followed by a buffer solution that runs vertically from top to bottom through multiple layers of sponge-like paper. One layer of the paper is packed with a series of lab-made peptides, components of proteins in the bacteria that causes Lyme disease, each of which detects a unique set of antibodies formed in response to that microbe. The patterns formed on the paper contain information about the presence and quantity of each antibody, which are captured by a digital reader and analyzed by an AI algorithm that delivers the results. In the study, each test paper cost $3, and the reader was a modified version of a commercially available $200 smartphone.
This new Lyme disease test is one of the first examples of a rapid diagnostic technology that comprehensively profiles a person's immune response to the disease.
“If we can quantify a panel of indicators from a single sample, it can tell us a lot of interesting things about a patient's condition,” said co-corresponding author Aydgan Ozkan, the Borgenau Professor of Engineering and Innovation at UCLA Samueli and CNSI's associate director for entrepreneurship, industry and academic exchange. “In Lyme disease, we're looking at a panel of immune factors that can vary widely from patient to patient, depending on their background, where they're from, etc. We needed AI to make sense of these complex signals.”
The researchers trained their algorithm using multiple samples, including those from early-stage disease, provided by the Bay Area Lyme Foundation's Lyme Disease Biobank, and then performed a blinded analysis of the technology's performance using new samples from the biobank. They achieved a sensitivity of 95.5% for detecting Lyme disease and a specificity of 100% for filtering out disease-free samples. Using additional samples from the CDC, they showed that their point-of-care test agreed highly closely with lab-based test results and could correctly detect Lyme disease and distinguish it from look-alike illnesses.
“AI is only as good as the data,” Ozkan says, “which is why we worked with the Lyme Biobank and the CDC to get well-characterized samples. Their collaboration was critical to helping our AI learn patterns of immune response to the bacteria that causes Lyme disease.”
In addition to AI, synthetic peptides from Connecticut-based BioPeptide Corp. were a key component of the new test technology. Compared to the whole proteins used in Lyme disease tests, peptides bind more specifically, are easier to manufacture and are more stable.
“Peptides are really important,” Di Carlo says. “We want to focus on a response that's very specific to Lyme disease, and not other related or similar diseases. At the same time, the test is cheaper, lasts longer, and has fewer missed diagnoses.”
With further success, it may be several years before the test can be introduced into the clinic. The researchers are currently looking for partners to scale the technology to speed up that process. They are also working on adapting the test to use whole blood samples instead of serum, with plans to further simplify the test format and develop a dedicated AI sample reader that is not dependent on a specific smartphone.
The study's co-first authors are Rajesh Ghosh, a graduate student in Di Carlo's lab at UCLA, and Hyoh Arm Jaun, a postdoctoral researcher in Ozkan's lab. Other authors are UCLA's Artem Goncharov, Bharath Palanisamy, Kevin Ngo, Katarina Pejcinovic, Nicole Crockenberger, and Omai Garner, Elizabeth Horne of the Lyme Disease Biobank, Ezdehar Ghazal and Andrew Okra of New York Medical College, and Paul Arnaboldi and Raymond Duttweiler of New York Medical College and Biopeptides, Inc.
The research was supported by the National Institutes of Health and the National Science Foundation.
