Analysis and Validation of Critical Signatures and Immune Cell Infiltration Characteristics in Doxorubicin-Induced Cardiotoxicity by Integrating Bioinformatics and Machine Learning

Abstract

Chao Huang,1,* Jixiang Pei,1,* Daisong Li,1 Tao Liu,2 Zhaoqing Li,1 Guoliang Zhang,1 Ruolan Chen,1 Xiaojian Xu,1 Bing Li,3,4 Zhexun Lian,1 Xian-Ming Chu1,5 1Department of Cardiology, the Affiliated Hospital of Qingdao University, Qingdao, Shandong, 266100, People’s Republic of China; 2The Affiliated Qingdao Central Hospital of Qingdao University, The Second Affiliated Hospital of Medical College of Qingdao University, Qingdao, Shandong, 266042, People’s Republic of China; 3Department of Genetics and Cell Biology, Basic Medical College, Qingdao University, Qingdao, 266000, People’s Republic of China; 4Department of Dermatology, The Affiliated Haici Hospital of Qingdao University, Qingdao, 266033, People’s Republic of China; 5The Affiliated Cardiovascular Hospital of Qingdao University, Qingdao, 266071, People’s Republic of China*These authors contributed equally to this workCorrespondence: Zhexun Lian; Xian-Ming Chu, Tel [phone], Email [email protected]; [email protected]: Doxorubicin-induced cardiotoxicity (DIC) is a severe side reaction in cancer chemotherapy that greatly impacts the well-being of cancer patients. Currently, there is still an insufficiency of effective and reliable biomarkers in the field of clinical practice for the early detection of DIC. This study aimed to determine and validate the potential diagnostic and predictive values of critical signatures in DIC.Methods: We obtained high-throughput sequencing data from the GEO database and performed data analysis and visualization using R software, GO, KEGG and Cytoscape. Machine learning methods and weighted gene coexpression network (WGCNA) were used to identify key genes for diagnostic model construction. Receiver operating characteristic (ROC) analysis and a nomogram were used to assess their diagnostic values. A multiregulatory network was built to reveal the possible regulatory relationships of critical signatures. Cell-type identification by estimating relative subsets of RNA transcript (CIBERSORT) analysis was used to investigate differential immune cell infiltration. Additionally, a cell and animal model were constructed to investigate the relationship between the identified genes and DIC.Results: Among the 3713 differentially expressed genes, three key genes (CSGALNACT1, ZNF296 and FANCB) were identified. A nomogram and ROC curves based on three key genes showed excellent diagnostic predictive performance. The regulatory network analysis showed that the TFs CREB1, EP300, FLI1, FOXA1, MAX, and MAZ modulated three key genes. An analysis of immune cell infiltration indicated that many immune cells (activated NK cells, M0 macrophages, activated dendritic cells and neutrophils) might be related to the progression of DIC. Furthermore, there may be various degrees of correlation between the three critical signatures and immune cells. RT‒qPCR demonstrated that the mRNA expression of CSGALNACT1 and ZNF296 was significantly upregulated, while FANCB was significantly downregulated in DOX-treated cardiomyocytes in vitro and in vivo.Conclusion: Our study suggested that the differential expression of CSGALNACT1, ZNF296 and FANCB is associated with cardiotoxicity and is also involved in immune cell infiltration in DIC. They might be potential biomarkers for the early occurrence of DIC.Keywords: doxorubicin, cardiotoxicity, biomarker, machine learning, immune infiltration

Keywords

• doxorubicin
• cardiotoxicity
• biomarker
• machine learning
• immune infiltration