A method for predicting natriuretic peptides in congenital heart disease using support vector machine

Atul Tyagi1, Sudeep Roy1, Ivo Provaznik2
1BUT Brno, 2Brno University of Technology


Abstract

Congenital heart disease (CHD) includes structural abnormalities of the heart that occur before birth. Congenital heart defects happen during the first eight weeks of the fetus development. CHD affected 0.8% of live births in the past few decades. The discovery of cardiac natriuretic peptides (NPs) such as ANP, BNP, CNP and research on their function and regulation in health and disease has led to breakthroughs in more profound understanding and clinical management of heart failure. Among NP properties are inhibition of cardiac remodeling. In cardiology, NPs are a valuable biomarker of heart failure. Studies investigating NP levels in the fetuses are quite limited. However, recent findings suggest that elevated NP levels are mainly attributed to increased central venous pressure secondary to arrhythmia caused by CHD. The features of ANP, BNP, and their related peptides in the umbilical cord blood and amniotic fluid provide a potential basis for their use as biomarkers. In our recent study, we analyzed 182 natriuretic peptides obtained from the UniProt database to predict and classify these peptides using Support Vector Machine (SVM). The di-peptide amino acid composition model achieved an accuracy of 92.86%, with Matthews correlation coefficient of 0.86.