Identifying Key MicroRNA Signatures for Neurodegenerative Diseases With Machine Learning Methods

Abstract

Neurodegenerative diseases, including Alzheimer's disease (AD), Parkinson's disease, and many other disease types, cause cognitive dysfunctions such as dementia via the progressive loss of structure or function of the body's neurons. However, the etiology of these diseases remains unknown, and diagnosing less common cognitive disorders such as vascular dementia (VaD) remains a challenge. In this work, we developed a machine-leaning-based technique to distinguish between normal control (NC), AD, VaD, dementia with Lewy bodies, and mild cognitive impairment at the microRNA (miRNA) expression level. First, unnecessary miRNA features in the miRNA expression profiles were removed using the Boruta feature selection method, and the retained feature sets were sorted using minimum redundancy maximum relevance and Monte Carlo feature selection to provide two ranking feature lists. The incremental feature selection method was used to construct a series of feature subsets from these feature lists, and the random forest and PART classifiers were trained on the sample data consisting of these feature subsets. On the basis of the model performance of these classifiers with different number of features, the best feature subsets and classifiers were identified, and the classification rules were retrieved from the optimal PART classifiers. Finally, the link between candidate miRNA features, including hsa-miR-3184-5p, has-miR-6088, and has-miR-4649, and neurodegenerative diseases was confirmed using recently published research, laying the groundwork for more research on miRNAs in neurodegenerative diseases for the diagnosis of cognitive impairment and the understanding of potential pathogenic mechanisms.

Keywords: classification algorithm; expression pattern; feature selection; microRNA; neurodegenerative disease.

FIGURE 1

Analysis flowchart for this study, which consists of three main steps: 1) miRNA dataset collection; 2) filtering and ranking of miRNA features in the dataset using Boruta, mRMR, and MCFS; 3) determining the essential miRNA features and building the best classifiers and classification rules using IFS method with RF and PART algorithms.

FIGURE 2

Venn diagram to show top ten miRNA features obtained by mRMR and MCFS methods. Four miRNA features are commonly identified.

FIGURE 3

IFS curves with different number of features in RF and PART under the mRMR and MCFS feature lists. (A). With the mRMR feature list, RF reaches the highest point (MCC = 0.683) with the top 106 features, and PART obtains the highest MCC (0.359) when using the top 72 features. The RF with top 41 features also provides high performance (MCC = 0.587). (B). With the MCFS feature list, RF and PART reach the highest points (MCC = 0.681 and 0.360, respectively) at the top 106 and 89 features. The RF with top 31 features also yields high performance (MCC = 0.575).

FIGURE 4

Performance of the key RF and PART classifiers on each class based on mRMR (A) and MCFS (B) feature lists. AD, VaD, DLB, MCI, and NC stand for Alzheimer’s disease, Vascular dementia, Dementia with Lewy bodies, Mild cognitive impairment and Normal control, respectively.

FIGURE 5

Number of rules generated by the optimal PART classifiers based on mRMR and MCFS feature lists. AD, VaD, DLB, MCI, and NC stand for Alzheimer’s disease, Vascular dementia, Dementia with Lewy bodies, Mild cognitive impairment and Normal control, respectively.

FIGURE 6

Performance of the key RF and PART classifiers without SMOTE. (A). Classifiers obtained by using mRMR feature list; (B). Classifiers obtained by using MCFS feature list. AD, VaD, DLB, MCI, and NC stand for Alzheimer’s disease, Vascular dementia, Dementia with Lewy bodies, Mild cognitive impairment and Normal control, respectively.

FIGURE 7

Venn diagram to show top 41 miRNA features obtained by mRMR method and top 31 miRNA features obtained by MCFS method. Nineteen miRNA features are commonly identified.