Identification of a Qualitative Signature for the Diagnosis of Dementia With Lewy Bodies

Shu Zhou; Qingchun Meng; Lingyu Li; Luo Hai; Zexuan Wang; Zhicheng Li; Yingli Sun

doi:10.3389/fgene.2021.758103

Identification of a Qualitative Signature for the Diagnosis of Dementia With Lewy Bodies

Front Genet. 2021 Nov 19:12:758103. doi: 10.3389/fgene.2021.758103. eCollection 2021.

Authors

Shu Zhou^{1

2}, Qingchun Meng², Lingyu Li², Luo Hai², Zexuan Wang², Zhicheng Li¹, Yingli Sun²

Affiliations

¹ Institute of Biomedical and Health Engineering, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
² Central Laboratory, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital & Shenzhen Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Shenzhen, China.

Abstract

Background and purpose: Diagnosis of dementia with Lewy bodies (DLB) is highly challenging, primarily due to a lack of valid and reliable diagnostic tools. To date, there is no report of qualitative signature for the diagnosis of DLB. We aimed to develop a blood-based qualitative signature for differentiating DLB patients from healthy controls. Methods: The GSE120584 dataset was downloaded from the public database Gene Expression Omnibus (GEO). We combined multiple methods to select features based on the within-sample relative expression orderings (REOs) of microRNA (miRNA) pairs. Specifically, we first quickly selected miRNA pairs related to DLB by identifying reversal stable miRNA pairs. Then, an optimal miRNA pair subset was extracted by random forest (RF) and support vector machine-recursive feature elimination (SVM-RFE) methods. Furthermore, we applied logistic regression (LR) and SVM to build several prediction models. The model performance was assessed using the receiver operating characteristic curve (ROC) analysis. Lastly, we conducted bioinformatics analyses to explore the molecular mechanisms of the discovered miRNAs. Results: A qualitative signature consisted of 17 miRNA pairs and two clinical factors was identified for discriminating DLB patients from healthy controls. The signature is robust against experimental batch effects and applicable at the individual levels. The accuracies of the-signature-based models on the test set are 82.61 and 79.35%, respectively, indicating that the signature has acceptable discrimination performance. Moreover, bioinformatics analyses revealed that predicted target genes were enriched in 11 Go terms and 2 KEGG pathways. Moreover, five potential hub genes were found for DLB, including SRF, MAPK1, YWHAE, RPS6KA3, and KDM7A. Conclusion: This study provided a blood-based qualitative signature with the potential to be used as an effective tool to improve the accuracy of DLB diagnosis.

Keywords: dementia with Lewy bodies; feature selection; microRNA; relative expression ordering; signature.