Parkinson’s disease (PD) is growing more rapidly than any other neurological disease, which makes its early detection so important. Researchers have developed a new machine-learning tool that shows promise as a way of detecting the disease early.
A diagnosis of PD usually occurs when a person presents with the traditional symptoms: slowed movements, tremors, poor balance and coordination, and muscular rigidity.
But the onset of atypical symptoms such as fatigue, trouble sleeping, bladder or bowel problems, depression and/or anxiety, and loss of smell can predate traditional PD symptoms by many years. A reliable method of testing for biomarkers that leads to an early diagnosis of PD instead of waiting for the appearance of traditional symptoms would mean that treatment of the disease can begin earlier.
Now, researchers from the University of New South Wales Sydney, in collaboration with Boston University, have harnessed the power of machine learning to develop a tool that shows promise as an early detector of PD.
Machine learning is widely used to develop accurate models for disease prediction. And advanced machine learning methods, like neural networks, are a way of processing large volumes of data. However, to be effective, the machine learning algorithm needs to be taught using data that is not ‘noisy’. Metabolomics, the large-scale study of metabolites, can be problematic in this regard.
Many metabolites – by-products created when the body breaks down food, drugs, and chemicals – are correlated with other metabolites, some of which do not contribute significantly to disease prediction.
That’s why the researchers developed a new machine learning tool, the Classification and Ranking Analysis using Neural network generates Knowledge from Mass Spectrometry or CRANK-MS.
“[T]o figure out which metabolites are more significant for the disease versus control groups, researchers usually look at correlations involving specific molecules,” said J Diana Zhang, lead author of the study. “But here we take into account that metabolites can have associations with other metabolites – which is where machine learning comes in. With hundreds to thousands of metabolites, we’ve used computational power to understand what’s going on.”
The researchers obtained metabolomic data from the Spanish European Prospective Investigation into Cancer and Nutrition (EPIC), focusing on 39 patients who’d developed PD and ran it through CRANK-MS. After comparing PD patients to healthy patients, the researchers were able to identify unique metabolic combinations that could be early warning signs of the disease.
The beauty of using CRANK-MS is that the researchers could use unadulterated data, which simplified the process.
“Typically, researchers using machine learning to examine correlations between metabolites and disease reduce the number of chemical features first, before they feed it into the algorithm,” said William Donald, the study’s corresponding author. “But here we feed all the information into CRANK-MS without any data reduction right at the start. And from that, we can get the model prediction and identify which metabolites are driving the prediction the most, all in one step. It means that if there are metabolites which may potentially have been missed using conventional approaches, we can now pick those up.”
While CRANK-MS was able to analyze metabolites indicative of PD with an accuracy of up to 96%, the researchers understand that the study’s small sample size means that further studies are needed.
The researchers say that, in future, CRANK-MS could be used at the first sign of atypical symptoms to ensure the early diagnosis of PD or to rule it out. And the machine learning algorithm is publicly available for researchers who might want to use it.
“We’ve built the model in such a way that it’s fit for purpose,” Zhang said. “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. What’s exciting is that CRANK-MS can be readily applied to other diseases to identify new biomarkers of interest.”
The study was published in the journal ACS Central Science.
Source: University of NSW Sydney
