Today at ADLM 2025 (formerly the AACC Annual Scientific Meeting & Clinical Lab Expo), researchers will unveil a blood test developed with the help of artificial intelligence (AI) that identifies Lyme disease sooner and more accurately than the current standard — and that could translate to vastly improved patient outcomes.
These findings spotlight the potential of AI to make a profound, positive difference in people’s lives when thoughtfully integrated into clinical laboratory medicine.
READ MORE: Researchers discover Achilles heel of Lyme disease pathogen
READ MORE: Alternative approach to Lyme disease vaccine development shows promise in pre-clinical models
Each year, more than 475,000 Americans are diagnosed with Lyme disease — a number that is only expected to climb due to climate change expanding the range of areas where ticks can live. When caught early, the condition responds well to antibiotics. However, the typical test — called two-tier serology — detects early Lyme accurately only 30% of the time. It’s a significant missed opportunity, since more than half of Lyme patients not diagnosed or treated within the first few weeks of infection will develop long-term health problems such as fatigue, neurocognitive issues, and arthritis.
Sensitivity and specificity
The new test leverages AI to offer major improvement. Its sensitivity and specificity are both over 90%, “meaning 9 out of 10 patients are going to get a correct diagnosis and receive appropriate treatment, which lowers the risk of chronic illness significantly,” said Holly Ahern, a microbiologist and chief scientific officer at ACES Diagnostics.
Ahern and team built on research in rhesus macaque monkeys, whose immune response to the bacteria causing Lyme is similar to humans, to develop a panel that looks for 10 proteins (antigens) and is completed as a single test. This approach is an improvement over the two-tier method, which may require up to four tests.
Human blood samples
Next, they analyzed blood samples from humans, including 123 people with Lyme disease and 197 uninfected individuals, to test whether adding machine learning to the test could bolster performance by detecting unique immune patterns. “You and I might get infected by the same bacteria, but we might both produce different antibody responses to it,” said Ahern. “With these antigens matched with a decision-tree–based classifier, we can actually pick that up in each individual case.”
The team found an algorithm that improved accuracy across all disease stages, correctly flagging infection in over 90% of early cases (versus 27% with the standard method). They hope the test — which, according to Ahern, is relatively inexpensive and works on standard laboratory equipment — will be commercially available by the end of 2026.
No comments yet