Analysis of Parkinson’s data with SAS® machine learning
McGuirk and two other computer scientists (Wei Xiao and Saba Emrani, both formerly of SAS) used the PPMI project data and used multitask machine learning to develop a model that can predict the progression of the disease for new Parkinson’s patients.
The SAS model predicted disease progression by predicting test results using the measurements from the initial, baseline evaluation, and subsequent visits all in one model. The model also helped determine which known biomarkers were most important – and discovered new ones.
McGuirk, Xiao and Emrani decided to rely on a multitask learning regression framework. “In the past, researchers have used conventional machine learning models that require the creation of a separate model for each of the 11 patient visits. But Parkinson’s disease is progressive, ”McGuirk says. “The results seen at the first visit are related to the second, third and all subsequent visits. The PPMI project was unique in using a multitask learning regression model that looks at all data from all visits simultaneously. ”
Taking into account the relationships between multiple tasks, multitask learning improves the overall performance of the model. “Say the data from a visit was messy – the clinician was sloppy with taking the tests,” McGuirk says. “If you try to predict Parkinson’s disease based on one visit, you would get an inaccurate reading. When you can tie the forecast to models for multiple visits, the results will be much more accurate. “
Breakthrough results can lead to a cure
The model McGuirk’s team developed proved better able to predict disease progression on each of the 11 patient visits over the 4.5-year study period than conventional approaches. “With accurate predictions, doctors will be able to start treatment sooner when it can have a greater impact,” McGuirk says.
In addition, the team was able to validate some already established biomarkers and discover new ones that are best able to differentiate between healthy individuals and those with the disease.
Eg. McGuirk’s team found that higher total cholesterol indicated a lower risk of Parkinson’s disease, while a combination of RNA markers, plasma, brain imaging, CSF measurements, and various non-motor assessments were important in predicting Parkinson’s disease progression.
With these newly discovered biomarkers, it is possible that Parkinson’s disease researchers can accelerate the development of effective treatments – and possibly even a cure for long-standing patients, such as Michael J. Fox. “That’s the goal,” McGuirk says.