A saliva test that detects epilepsy, Parkinson's, and schizophrenia with up to 98% accuracy

Diagnosing neurological disorders typically requires expensive PET scans or invasive cerebrospinal fluid draws. A team of Korean researchers has demonstrated a different approach: a platform that can classify epilepsy, Parkinson's disease, and schizophrenia from a small sample of saliva, with accuracy reaching 98%.

The technology, published in Advanced Materials, was developed jointly by Dr. Sung-Gyu Park at the Korea Institute of Materials Science (KIMS), Prof. Ho Sang Jung's team at Korea University, and researchers from the College of Medicine at The Catholic University of Korea.

How nanostructures read protein shape

The system is built around a principle called surface-enhanced Raman spectroscopy (SERS), which uses the interaction between light and molecular vibrations to produce a spectral fingerprint of whatever molecule sits on a specially designed surface. The challenge has always been signal strength: the Raman signals produced by biological molecules are vanishingly weak.

The Korean team solved this with a platform they call Galvanic Molecular Entrapment (GME)-SERS. The device uses nanostructures composed of copper oxide and gold (Au-CuO) that create plasmonic "hotspots" -- nanoscale regions where electromagnetic fields are concentrated so intensely that they amplify Raman signals by more than a billion times. When proteins from saliva are captured on these structures, their spectral signatures become readable.

What sets this approach apart from conventional diagnostics is what it measures. Standard blood tests typically quantify protein concentration -- how much of a given protein is present. The GME-SERS platform instead detects structural changes in proteins, specifically distinguishing between normal protein forms (monomers) and abnormally aggregated forms (fibrils). Protein fibrillation is a hallmark of several neurological conditions, including Parkinson's disease, but has been extremely difficult to detect with conventional techniques.

Testing on real patients

The research team analyzed saliva samples from 44 patients diagnosed with epilepsy, schizophrenia, or Parkinson's disease, along with 23 healthy controls, in collaboration with St. Vincent's Hospital. The platform classified these disorders with accuracy exceeding 90%, reaching up to 98% for some conditions.

The ability to distinguish neurological disorders based on fundamental pathological indicators -- structural protein changes rather than total protein counts -- is what the researchers describe as the key advance. Protein misfolding and aggregation are among the earliest molecular events in many brain diseases, potentially appearing before clinical symptoms become obvious.

From hospital labs to portable devices

"An era has begun in which brain disease conditions can be assessed through simple saliva analysis without the need for costly PET imaging or cerebrospinal fluid testing," said Dr. Park. Prof. Jung added that the technology's non-invasive, low-cost nature gives it significant potential for expansion beyond hospital settings to home-based diagnostic devices.

The research team plans to pursue commercialization through portable Raman sensor-based point-of-care devices and technology transfer to medical and life-science companies.

What remains uncertain

The study's sample size of 67 individuals across four groups (three diseases plus controls) is small by diagnostic validation standards. Clinical diagnostics typically require testing across hundreds or thousands of patients before a technology can be considered reliable enough for routine use. The high accuracy figures, while encouraging, need replication in larger and more diverse populations.

It is also not yet clear how the platform performs in early-stage disease, when symptoms are mild and protein structural changes may be subtle. The patients in this study had established diagnoses, meaning the test was distinguishing known disease from known health rather than detecting disease in an unscreened population -- a much harder task.

The overlap between conditions also warrants attention. Protein structural changes are not unique to a single disease, and how reliably the platform can differentiate between conditions that share molecular features will need further testing. Whether saliva-based protein profiles remain stable across different times of day, hydration states, and other variables that affect saliva composition has not been addressed.

Still, as a proof of concept, the work is notable. Saliva collection requires no needles, no imaging equipment, and no trained phlebotomists. If the accuracy holds in larger studies, a diagnostic that requires nothing more than spitting into a cup would be a meaningful advance for neurological medicine.