Predictive Biomarkers for Postmyocardial Infarction Heart Failure Using Machine Learning: A Secondary Analysis of a Cohort Study. (13th December 2021)

• Record Type: Journal Article
• Authors: Li, Feng; Sun, Jin-Yu; Wu, Li-Da; Qu, Qiang; Zhang, Zhen-Ye; Chen, Xu-Fei; Kan, Jun-Yan; Wang, Chao; Wang, Ru-Xing
• Other Names: Liang Zhaohui Academic Editor.
• Is Part Of: Evidence-based complementary and alternative medicine. Volume 2021(2021)
• Publication Date: 2021-12-13
• Subjects: Alternative medicine -- Periodicals; 615.505
• DOI: 10.1155/2021/2903543
• Languages: English
• ISSNs: 1741-427X

Abstract

Abstract : Background . There are few biomarkers with an excellent predictive value for postacute myocardial infarction (MI) patients who developed heart failure (HF). This study aimed to screen candidate biomarkers to predict post-MI HF. Methods . This is a secondary analysis of a single-center cohort study including nine post-MI HF patients and eight post-MI patients who remained HF-free over a 6-month follow-up. Transcriptional profiling was analyzed using the whole blood samples collected at admission, discharge, and 1-month follow-up. We screened differentially expressed genes and identified key modules using weighted gene coexpression network analysis. We confirmed the candidate biomarkers using the developed external datasets on post-MI HF. The receiver operating characteristic curves were created to evaluate the predictive value of these candidate biomarkers. Results . A total of 6, 778, 1, 136, and 1, 974 genes (dataset 1) were differently expressed at admission, discharge, and 1-month follow-up, respectively. The white and royal blue modules were most significantly correlated with post-MI HF (dataset 2). After overlapping dataset 1, dataset 2, and external datasets (dataset 3), we identified five candidate biomarkers, including FCGR2A, GSDMB, MIR330, MED1, and SQSTM1 . When GSDMB and SQSTM1 were combined, the area under the curve achieved 1.00, 0.85, and 0.89 in admission, discharge, and 1-month follow-up, respectively. Conclusions . This study demonstrates that FCGR2A, GSDMB, MIR330, MED1, and SQSTM1 are the candidate predictive biomarker genes for post-MI HF, and the combination of GSDMB and SQSTM1 has a high predictive value.