Machine learning Ensemble for the Parkinson’s disease using protein sequences

Authors: Arora, P., Mishra, A. and Malhi, A.

Journal: Multimedia Tools and Applications

Volume: 81

Issue: 22

Pages: 32215-32242

eISSN: 1573-7721

ISSN: 1380-7501

DOI: 10.1007/s11042-022-12960-7

Abstract:

The most challenging issue in diagnosing and treating neurological disorders is gene identification that causes the disease. Classification of the genes that cause or initiate different genes leading to diseases with neurological disorders like Parkinson’s disease, is a grave challenge in biomedical research. Detecting neurological disorders has a significant contribution in genetics, which require the deployment of machine learning methods which are still in its infancy. For exploring protein sequences (genes), computational analysis is a vital technique, since a manual comparison of multiple sequences results in impracticality. It helps to find a gene in the sequence and to combine protein sequences into a class of similar sequence. There are many traditional methods available for detection of Parkinson’s disease. However, relatively less comparative work is done on Machine Learning techniques using protein sequences for evaluation of Parkinson’s disease. This paper demonstrates the comparison of multiple classification methods to identify Parkinson’s disease using hydrophobicity and Amino Acid Composition as feature extraction methods. Classification methods are then combined to propose a 2-level ensemble method based on the false prediction rate. The performance of these methods is evaluated using Precision, Recall, F-Score and ROC curves. Experimental results have demonstrated that Random Forest, SVM, Neural Network (PCANNET) and Naïve Bayes classifiers individually performed best based on their performance metrics under 5-fold cross validation, whereas the proposed method outperforms Random Forest and SVM by 1.96%, NB by 1.1% and PCANNET by 1.68% respectively. Further, statistical analysis has been added to validate the proposed method.

Source: Scopus