Abstract P051: Application Of Artificial Intelligence In Transcriptome-based Diagnosis Of Cardiomyopathies

• Published in: Hypertension (Dallas, Tex. 1979) Tex. 1979), 2020-09, Vol.76 (Suppl_1)
• Main Authors: Cheng, Xi, Alimadadi, Ahmad, Manandhar, Ishan, Aryal, Sachin, Munroe, Patricia B, Joe, Bina
• Format: Article
• Language: English
• ISSN: 0194-911X; 1524-4563
• DOI: 10.1161/hyp.76.suppl_1.P051

Abstract only Dilated cardiomyopathy (DCM) and ischemic cardiomyopathy (ICM) are predisposing conditions for increased risk of cardiovascular diseases including heart failure, valve disease and arrhythmias. An accurate diagnostic strategy of differentially classifying different types of cardiomyopathies could contribute to precision medicine in routine clinical care. We hypothesized that artificial intelligence (AI) could be trained with cardiac transcriptomic data for diagnostic classifications of clinical cardiomyopathies. To test this hypothesis, various supervised machine learning (ML) models, such as support vector machine with radial kernel (svmRadial), neural networks with principal component analysis (pcaNNet), decision tree (DT), elastic net (ENet) and random forest (RF), were used to analyze RNA-seq data of human left ventricle tissues collected from 41 DCM patients, 47 ICM patients and 49 non-failure controls (NF). Initial ML classifications achieved an AUC of ~0.96 (ENet, pcaNNet and svmRadial) for NF vs DCM, an AUC of ~0.89 (RF) for NF vs ICM, and an AUC of ~0.90 (svmRadial) for DCM vs ICM. Next, 50 highly contributing genes (HCGs) for classifying NF and DCM, 68 HCGs for classifying NF and ICM, and 59 HCGs for classifying DCM and ICM were selected for re-training ML models. The re-trained models achieved an AUC of ~0.96 (RF) for NF vs DCM, an AUC of ~0.97 (pcaNNet) for NF vs ICM, and an AUC of ~0.94 (RF) for DCM vs ICM. Pathway analyses of the selected HCGs further demonstrated their pathophysiological roles in cardiovascular dysfunctions including cardiomyopathies. Overall, our study demonstrates the promising potential of using AI via ML models as a novel approach to achieve a greater level of precision in diagnosing different types of clinical cardiomyopathies.