Objective: Heart failure with preserved ejection fraction (HFpEF) is a disease associated with significant clinical pathophysiological heterogeneity in which maladaptive cardiac fibrosis, in both the right and left ventricles of the heart, plays a unique role in the manifestation of disease. The prevalence of HFpEF is greater in postmenopausal women with hypertension. Thus, the purpose of this study was to identify variables predictive of cardiac fibrosis with the potential to help elucidate new biomarkers indicative of the pathological diffuse fibrotic remodeling often observed in HFpEF. Given significant sex disparities noted in HFpEF prevalence, biological inputs were measure in both the right and left ventricles of a pre-clinical female Yucatan mini-swine model of chronic pressure overload-induced heart failure. Predictors of total left ventricular collagen content were identified and assessed for accuracy.
Hypothesis: Artificial neural networks can identify molecular markers involved in the bioregulation of the extracellular matrix (ECM) that predict total collagen content in the left ventricle with gt;90% accuracy.
Methods: 98 biological features measured in the right and left ventricle including mRNA, protein, and activity measurements of estrogen/progesterone receptors, ECM regulatory biomarkers including matrix metalloproteinases (MMP’s) and their tissue-inhibitors (TIMP’s), several aspects of the mitogen-activated protein kinase signaling pathway including ERK, JNK, and dual specificity phosphatases, and ECM structural components (e.g. collagen, fibronectin) were analyzed in 24 female Yucatan mini-swine. Variables were pooled and assessed comprehensively from animals divided into 4 groups: control, ovariectomized, aortic-banded, and aortic-banded with ovariectomy (n=5-7/group). Missing data were mean imputed and the min-max normalization method was used. Linear regression models were used to identify variables with the strongest relationships to left ventricular total collagen content. Identified molecular predictors were used in an artificial neural network model using stratified 4-fold cross validation with logistic activation function, one hidden layer, and 4 nodes.
Results: In the left ventricle, TIMP2 protein and mRNA, MMP2 mRNA, and fibronectin protein and mRNA were used in the artificial neural network. In the right ventricle, MMP9 activity/abundance (zymography) and total collagen content were included in the computational analysis. Together, these variables could predict total collagen content in the left ventricle with 92.08 % accuracy.
Conclusion: Analysis of cardiac fibrosis by artificial neural network identified seven bi-ventricular molecular features with a high capacity to predict total collagen content in the left ventricle. These results highlight their potential as important biomarkers of pathological cardiac fibrotic remodeling in a preclinical setting of chronic pressure overload-induced heart failure with potential relevance to HFpEF.
