AI-Powered Blood Test Accurately Classifies Multiple Neurodegenerative Diseases and Detects Co-Existing Conditions
核心洞察
Researchers at WashU Medicine developed GPND-AI, an AI classifier that analyzes 15 blood proteins to distinguish Alzheimer's, Parkinson's, frontotemporal dementia (搜索), and dementia with Lewy bodies (搜索) with over 90% accuracy.
The tool uniquely detects mixed pathologies, identifying when patients harbor multiple neurodegenerative disease processes simultaneously rather than forcing a single diagnosis.
External validation using autopsy-confirmed cases demonstrated the test's biological validity, with classifier outputs aligning closely with actual brain tissue pathology.
A research team at Washington University School of Medicine in St. Louis (搜索) (WashU Medicine) has developed an artificial intelligence-powered blood test capable of accurately distinguishing among several major neurodegenerative diseases and detecting when multiple conditions co-exist in a single patient. The classifier, called GPND-AI (Generalizable Protein-based Neurodegenerative Disease Artificial Intelligence), analyzes 15 specific proteins from a standard blood draw and demonstrated a diagnostic accuracy rate of over 90% in distinguishing Alzheimer's disease (搜索), Parkinson's disease (搜索), frontotemporal dementia (搜索), and dementia with Lewy bodies (搜索) from each other and from normal cognitive aging.
The findings, published in Alzheimer's & Dementia, the journal of the Alzheimer's Association, represent a significant step toward accessible, objective diagnostic tools for a disease category that currently affects over 57 million people worldwide.
Overcoming Diagnostic Ambiguity
Diagnosing neurodegenerative diseases correctly remains a persistent challenge in clinical practice. Patients exhibiting signs of cognitive decline often navigate a time-consuming and costly clinical journey filled with uncertainty. Symptoms of different dementias frequently overlap, making clinical evaluations subjective and prone to human error. Physicians struggle to determine whether a patient has Alzheimer's disease (搜索) or another form of dementia based solely on memory tests, motor function evaluations, and behavioral assessments.
While invasive procedures such as lumbar punctures or expensive imaging scans like MRI and PET offer clarity, these methods remain inaccessible for many patients due to high systemic costs and limited regional availability. Rural populations and individuals without comprehensive health insurance often face significant financial and geographical barriers when seeking specialized neurological evaluations.
Recognizing the urgent need for an accessible diagnostic alternative, the WashU Medicine research group aimed to bridge this gap by developing a non-invasive tool capable of reflecting the biological complexity of the aging brain. They hypothesized that analyzing blood plasma could democratize expert-level neurological diagnostics across diverse medical settings.
The 15-Protein Biomarker Panel
To build the diagnostic tool, researchers analyzed proteomic and clinical information sourced from the Charles F. and Joanne Knight Alzheimer's Disease (搜索) Research Center and the Movement Disorder Clinic. Using the NUcleic acid-Linked Immuno-Sandwich Assay (NULISA) central nervous system panel, the team examined thousands of circulating proteins in human blood. This assay technology enables high-sensitivity detection of low-abundance proteins in plasma.
Through rigorous machine learning protocols, the scientists narrowed molecular candidates to a core set of 15 informative proteins. These selected biomarkers reflect key elements of neurodegenerative pathology, including synapse damage, nerve deterioration, and brain inflammation. The targeted proteomics approach ensures the test captures the most vital physiological indicators of disease progression.
During the testing phase, the 15-protein panel achieved an area under the ROC curve of 0.955 and a 92.3% accuracy rate across five distinct diagnostic categories. To ensure reliability, the researchers performed external validation using a separate cohort from the Banner Sun Health Research Institute. The tool matched its initial success, proving the algorithm could generalize accurately across different patient populations. Critically, the classifier outputs aligned closely with the actual pathological burden discovered in brain tissue upon autopsy, cementing the biological validity of the blood test.
Detecting Mixed Pathologies
A distinct capability of GPND-AI lies in its ability to detect mixed pathologies. In everyday clinical settings, older patients frequently develop overlapping neurodegenerative conditions simultaneously. A person might show outward symptoms of Parkinson's disease (搜索) but simultaneously harbor Alzheimer's disease (搜索) pathology within their brain tissue. Traditional diagnostic frameworks force patients into a single disease category, ignoring secondary biological processes.
"Many patients get labeled with a single diagnosis of, say, Alzheimer's or Parkinson's, but in reality their brains often show a mixture of disease injuries," said Dr. Carlos Cruchaga, the senior author of the study and the Barbara Burton and Reuben M. Morriss III Professor in the Department of Psychiatry at WashU Medicine. He explained that existing medical tools lacked the specific design necessary to capture this intricate biological overlap. "Our goal was to build a test that doesn't just say 'yes' or 'no' to one disease but instead gives an indication of all the major neurodegenerative diseases happening in that person."
Implications for Clinical Trials and Precision Medicine
The ability to map co-existing neurodegenerative processes opens new possibilities for precision medicine in neurology. When physicians can identify the exact combination of protein biomarkers in a patient's blood, they can tailor clinical interventions to those specific biological drivers. This level of biological insight prevents the prescription of drugs that might accidentally exacerbate secondary conditions and helps neurological specialists anticipate how a patient's symptoms might evolve.
Beyond individual patient care, the blood-based classifier stands to accelerate biopharma research and drug development. Clinical trials testing new neurodegenerative therapies currently experience high failure rates, partially due to poor patient selection criteria. If a trial tests a drug designed to clear amyloid plaques but a large portion of participants have frontotemporal dementia (搜索) instead of Alzheimer's, the trial data will skew negatively and hide potential drug efficacy. The new AI tool allows clinical researchers to screen and enroll the right patients for specific disease pathways using a simple blood draw. Implementing this technology could save the healthcare system millions of dollars by eliminating redundant and expensive diagnostic imaging procedures during trial screening phases.
The Path Forward
While the 15-protein blood test represents a scientific advancement, the underlying technology requires further refinement before commercial release. Medical researchers must conduct additional prospective longitudinal studies to standardize blood testing protocols and secure regulatory approvals. Nevertheless, the foundational evidence provided by the WashU Medicine team proves the utility of proteomics-based approaches for complex dementia diagnostics. In the near future, this accessible blood test could become a routine part of senior healthcare, empowering patients, guiding specialized care, and accelerating the discovery of targeted therapies for brain disorders.