Objective: In Parkinson’s disease (PD) non-motor symptoms often appear before the motor symptoms. Detecting and understanding these non-motor symptoms and their possible genetic or epigenetic connections is important. Early identification of the disease can improve diagnosis and management, which can ultimately enhance the quality of life for those affected by this condition.

Background: PD is characterized by the degeneration of dopaminergic neurons, which leads to the onset of motor symptoms. To learn more about Parkinson’s disease, researchers have utilized the Parkinson’s Progression Markers Initiative (PPMI). This initiative is a crucial source of information that provides a wealth of data on both healthy individuals and patients with Parkinson’s. Our study aims to build machine learning based classification model of the DNA methylation and non-motor symptom assessments in PD, by utilizing data available in PPMI.

Method: The assessment score of each non-motor symptom (or features) has been categorized into three categories, viz. Normal, Mild, and Severe. Feature selection were carried out (feature common in at least two methods) using four different algorithms, viz. Decision Trees, Random Forest, Boruta and Normalized Mutual Information (NMI). Among various algorithms Random Forest performed better to identify the model accuracy. NMI was performed to check synergistic effect of any two features promotes the disease progression. Dimension reduction techniques performed on DNA methylation data. Important CpG and corresponding genes related to each feature were identified.

Results: Important features (eight in each case) found in male and female were different, smell disorder and cognitive impairment only being the common ones. Gender specific classification models show high accuracy (>82%). Top CpGs in each feature were selected using NMI and Boruta, and models have been prepared, and their methylation pattern (Hoehn Yahr Stage wise) shows perturbation in diseased and control sets.

Conclusion: Smell loss and cognitive impairment are perturbed commonly in male and female while other features are different, which suggest separate set of tests (questionnaires) can be set for disease diagnosis in male and female. The models built can be used as peri-diagnostic tool or along with established clinical tests to classify disease from healthy ones.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/gender-specific-classification-models-for-parkinsons-disease-using-non-motor-symptoms-and-dna-methylation-data/