In a groundbreaking study published in ACS Central Science, researchers have leveraged the power of neural networks to analyze biomarkers in patients’ bodily fluids, paving the way for early detection and potential prevention of Parkinson’s disease. Led by scientists from the UNSW School of Chemistry, the team utilized blood samples from healthy individuals collected by the Spanish European Prospective Investigation into Cancer and Nutrition (EPIC) to develop a machine learning tool called CRANK-MS (Classification and Ranking Analysis using Neural network generates Knowledge from Mass Spectrometry).

Traditionally, metabolomics data analysis has relied on statistical approaches and limited correlations involving specific molecules. However, the researchers adopted a different approach by considering associations between metabolites and harnessing the computational power of CRANK-MS to understand the complex interactions within the data. By examining extensive metabolite datasets from both Parkinson’s patients and matched control subjects, they identified unique combinations of metabolites that could serve as potential indicators or early warning signs of Parkinson’s disease.

Unlike conventional methods that reduce the number of chemical features before feeding data into the algorithm, CRANK-MS processed the unedited dataset in its entirety. This holistic approach allowed the researchers to make model predictions and determine the metabolites driving the prediction without data reduction. By avoiding data reduction, they uncovered metabolites that may have been overlooked using traditional approaches. These findings represent a significant step toward developing a non-invasive diagnostic tool for Parkinson’s disease, which currently relies on physical symptom observation.

Parkinson’s disease diagnosis typically occurs after the manifestation of motor symptoms, but atypical symptoms such as sleep disorders and apathy can precede them by decades. CRANK-MS holds the potential to detect Parkinson’s at the first sign of atypical symptoms, enabling early intervention and personalized care. However, it’s important to note that validation studies on larger cohorts conducted across multiple regions are necessary before the tool can be reliably used in clinical practice.

The study yielded promising results, with CRANK-MS demonstrating an accuracy of up to 96% in detecting Parkinson’s disease by analyzing blood chemicals. This high accuracy in predicting Parkinson’s disease prior to clinical diagnosis is noteworthy. Moreover, the identification of chemical markers that play a crucial role in accurate prediction aligns with previous findings from cell-based assays. The study also unveiled intriguing insights, such as the lower concentrations of triterpenoids, known neuroprotectants found in foods like apples, olives, and tomatoes, in the blood of individuals who later developed Parkinson’s. This opens the door to investigating whether the consumption of these foods could offer natural protection against the disease. Additionally, the presence of polyfluorinated alkyl substances (PFAS) in those who developed Parkinson’s warrants further exploration as a potential link to exposure to industrial chemicals.

CRANK-MS, a publicly available tool, represents a significant advancement in disease diagnosis using metabolomics data. Its user-friendly interface and efficient processing time make it accessible to researchers, generating results in under 10 minutes on a conventional laptop. The application of AI and machine learning in disease diagnosis holds immense promise, not only for Parkinson’s but also for other diseases, as CRANK-MS can identify new biomarkers of interest.

The researchers’ groundbreaking work demonstrates the potential of AI to revolutionize disease detection and monitoring, ultimately leading to more effective and personalized healthcare. With further refinement and validation, CRANK-MS could contribute to a paradigm shift in early diagnosis and intervention strategies, improving outcomes for patients worldwide.

By Impact Lab