AUTHOR=Zhang Lin , Lin Yexiang , Wang Kaiyue , Han Lifeng , Zhang Xue , Gao Xiumei , Li Zheng , Zhang Houliang , Zhou Jiashun , Yu Heshui , Fu Xuebin 

TITLE=Multiple-model machine learning identifies potential functional genes in dilated cardiomyopathy

JOURNAL=Frontiers in Cardiovascular Medicine

VOLUME=Volume 9 - 2022

YEAR=2023

URL=https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2022.1044443

DOI=10.3389/fcvm.2022.1044443

ISSN=2297-055X

ABSTRACT=Machine learning (ML) has been widely performed in diagnosing diseases. Filtering the variables with a single ML to might miss several essential genes. We developed a novel overall normalized sum weight of multiple MLs to assess whether genes can be conducive to diagnosing Dilated cardiomyopathy (DCM) and explore its correlation with immune regulation. A total of six eligible microarrays (386 sample size) were included in this study. With the ratio of  5:1, two were involved in a training set, and four were in the testing set. Fifty-five differently expressed genes (DEGs) were identified in the control and DCM individuals (29 up-regulated and 26 down-regulated). Furthermore, we developed a normalized sum weight of six MLs methods, including Support Vector Machine (SVM), Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest (RF), Gradient Boosting Machine (GBM), Decision Trees (DT), and Neural Network (NN) to identify potential candidate genes (overall weights >1). Consequently, ten genes were selected for model establishment. According to the receiver operating characteristic (ROC), three genes, SERPINA3, FRZB, and FCN3, were finally identified. The average values of area under the curve (AUC) for these three genes are higher than 0.85 in both training and testing sets. Significantly, the ROC of SERPINA3 was >0.9 in both training and testing sets. Besides, these three genes also significantly correlated to five immune cells, namely Monocytes, T cells CD8, T cells CD4 memory resting, Plasma cells, and T cells regulatory (Tregs) (P<0.05).