Interpretable machine learning on metabolomics data reveals biomarkers for Parkinson’s disease

The use of machine learning (ML) with metabolomics provides opportunities for the early diagnosis of disease. However, the accuracy and extent of information obtained from ML and metabolomics can be limited owing to challenges associated with interpreting disease prediction models and analysing many chemical features with abundances that are correlated and ‘noisy’. Here, we report an interpretable neural network (NN) framework to accurately predict disease and identify significant biomarkers using whole metabolomics datasets without feature selection. The performance of the NN approach for predicting Parkinson’s disease (PD) from blood plasma metabolomics data was significantly higher than classical ML methods with a mean area under the curve of > 0.995. PD-specific markers that contribute significantly to early disease prediction were identified including an exogenous polyfluoroalkyl substance. It is anticipated that this accurate and interpretable NN-based approach can improve diagnostic performance for many other diseases using metabolomics and other untargeted ‘omics methods.