Parkinson’s disease (PD) progresses more rapidly than other neurological disorders, so early detection is very important. Researchers have developed a new machine-learning tool that shows promise for early detection of disease.
A diagnosis of PD is usually made when traditional symptoms such as slowness of movement, tremors, poor balance and coordination, and muscle stiffness appear.
However, the onset of atypical symptoms such as fatigue, sleep disturbances, bladder and bowel problems, depression and anxiety, and loss of smell can predate traditional PD symptoms by many years. A reliable method of testing for biomarkers leading to an early diagnosis of PD, rather than waiting for the traditional symptoms to appear, would allow treatment of the disease to begin sooner.
Now, researchers at the University of New South Wales, Sydney, in collaboration with Boston University, have harnessed the power of machine learning to develop a tool that promises to be an early detector of PD.
Machine learning is widely used to develop accurate models for predicting disease. Also, advanced machine learning techniques such as neural networks are ways to process large amounts of data. However, to be effective, machine learning algorithms must be trained using “noisy” data. Metabolomics, large-scale studies of metabolites, can pose problems in this respect.
Many metabolites (byproducts produced when the body breaks down foods, drugs, and chemicals) are correlated with other metabolites, some of which are less predictive of disease.
That’s why researchers developed a new machine learning tool: classification and ranking analysis using neural networks that generate knowledge from mass spectrometry or CRANK-MS.
“[T]When trying to figure out which metabolites are more important for disease compared to controls, researchers typically look at correlations involving specific molecules,” said J Diana Zhang, lead author of the study. Stated. “But here we take into account that metabolites can be related to other metabolites. This is where machine learning comes in. We can use hundreds or thousands of metabolites to figure out what I used my computational power to understand what was going on.”
The researchers obtained metabolomic data from the European Prospective Investigation on Cancer and Nutrition (EPIC) in Spain, focusing on 39 patients who developed PD and underwent CRANK-MS. By comparing PD patients with healthy patients, the researchers were able to identify unique metabolic combinations that may be early warning signs of the disease.
The advantage of using CRANK-MS is that it simplifies the process as researchers can use pure data.
“Typically, researchers using machine learning to explore metabolite-disease correlations first reduce the number of chemical features before feeding them into the algorithm,” said William Donald, corresponding author of the study. said Mr. “But here we don’t do any data reduction to begin with, we just feed all the information into CRANK-MS. And from there we get the model predictions, and we all know which metabolites are driving the predictions the most.” step, which means we can now detect metabolites that may have been missed by traditional approaches.”
Although CRANK-MS was able to analyze metabolites indicative of PD with up to 96% accuracy, the researchers understand that the small sample size of this study warrants further investigation.
In the future, the researchers say, CRANK-MS may be used at the first sign of atypical symptoms to ensure early diagnosis or rule out PD. there is Also, the machine learning algorithms are publicly available for researchers who wish to use them.
“We built the model in a fit-for-purpose manner,” says Zhang. “The application of CRANK-MS to detect Parkinson’s disease is just one example of how AI can improve the way we diagnose and monitor disease. It can be easily applied to identify new biomarkers of interest.”
The study was published in a journal ACS Central Science.
Source: University of NSW Sydney
