A new blood test for Lyme disease outperforms standard diagnostic testing across all stages of the disease, according to new data presented at the 2025 Association for Diagnostics and Laboratory Medicine annual meeting in Chicago.
The test, which pairs a multiplexed assay with a machine learning algorithm, showed over 90% accuracy in diagnosing Lyme disease in 308 human serum samples.
The CDC estimates that about 476,000 individuals may be diagnosed and treated for Lyme disease each year in the US. The tick-borne disease, caused primarily by infection with Borrelia burgdorferi, is most treatable in the first few weeks of infection, but diagnosis during this early stage poses a challenge to clinicians.
Barriers to Early Diagnosis
The most common sign of early infection is a skin rash called erythema migrans (EM) that often occurs at the site of the tick bite. While a bull-eye-shaped rash is telltale sign of Lyme disease, this type of EM is actually not a common presentation, said Liz Horn, PhD, MBI, principal investigator of the Lyme Disease Biobank, a biorepository of samples from people with Lyme disease, headquartered in Portland, Oregon. She was not involved with the work. EMs can present in different forms, and about 30% of people with Lyme disease do not develop a rash.
“If you don’t have a rash, you have to rely on serologic tests,” Horn said, which are notoriously insensitive in early disease, as antibodies can take weeks to develop.
The CDC currently recommends a two-test format. Standard two-tier testing (STTT) involves an enzyme immunoassay as the first step, which, if positive or indeterminate, is followed by a Western Blot. In the first few weeks of disease, antibody levels can be too low to detect, leading to false negatives.
In a recent study led by Horn using samples from individuals with signs and symptoms of early Lyme disease who were enrolled in the Lyme Disease Biobank, just 34% of patients presenting with EM > 5 cm tested positive on serologic tests. Among patients without EM, only 16% had positive results, and among those with EM ≤ 5 cm, the positivity rate dropped to just 5%.
“If we could get a serology test that works better in early disease, that could have great benefits for both patients and providers,” Horn said.
Pairing AI With a Multiassay
This new blood test, called LymeSeek, works differently by detecting 10 different antigens simultaneously. An algorithm developed using deep learning “sorts through all of the responses,” explained Holly Ahern, MS, MT, the chief scientific officer of ACES Diagnostics, the company developing LymeSeek, in Paradise Valley, Arizona. “That is what gives the test its very high sensitivity and specificity,” she said. Ahern is also an associate professor of microbiology at State University of New York at Adirondack (SUNY Adirondack) in Queensbury, New York.
To validate the test, developers analyzed the human serum samples of 150 control individuals and 158 Lyme disease cases, across all phases of Lyme disease. Samples were provided by the Lyme Disease Biobank, the CDC, and the Johns Hopkins University Lyme Disease Research Center in Baltimore. The control samples were from patients in endemic and nonendemic areas and included individuals with non-Lyme look-like diseases, like mononucleosis and rheumatoid arthritis.
The test correctly classified 281 samples as Lyme cases or control individuals, resulting in a test sensitivity of 91.7% and a specificity of 90.7%.
In an additional analysis, the ACES diagnostics team also compared LymeSeek to STTT using samples from the Johns Hopkins Lyme Disease Research Center. In very early stages of disease (0-72 hours after appearance of an EM rash), LymeSeek diagnosed 100% of Lyme cases, whereas standard testing identified only 37% of cases. In samples taken 3-4 weeks after EM appearance, LymeSeek’s results remained consistent, and standard testing identified 76% of cases.
“The observed seroconversion indicates the enrolled patients did have Lyme disease and is a confirmation that the test result given by LymeSeek is a true positive test result,” Ahern added.
In patients with post-treatment Lyme disease (6 months or more after infection), LymeSeek identified 97% of cases, whereas standard testing identified less than half (46%).
“With this one single test, we can detect [Lyme] in the early stages, but we can also detect it across all stages of the disease” Ahern said, which includes through 1 year from first Lyme diagnosis.
ACES Diagnostics is now working to develop a clinical trial plan with the FDA. The test has received a De Novo classification, meaning it’s being evaluated on its own merits rather than compared to the current standard diagnostics.
“We’re not just showing that we are as good as two-tier serology because what’s the point of being as good as a test that really isn’t very good?” Ahern said.
Horn and Ahern had no relevant disclosures.