Machine learning to differentiate diseased cardiomyocytes from healthy control cells

Abstract

Abstract Human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) have been shown to be useful to improve techniques that are developed for the study of cardiac disease. Abnormalities in Ca2+ transients are commonly present in iPSC-CMs derived from individuals with a cardiac disease. We previously observed that Ca2+ transient signals of healthy CMs can be distinguished from transients of CMs derived from individuals having different genetic cardiac diseases. Machine learning was used to distinguish different diseases from each other as well as from controls. We wanted further to investigate whether we are able to separate iPSC-CM Ca2+ signals of any genetic cardiac disease as one group from those of healthy individuals by utilizing machine learning methods. A total number of 593 CM transient signals from healthy individuals and from patients were analyzed. We obtained a best classification accuracy of 87% between the disease group and controls. This finding provides evidence that machine learning methods are efficient for identifying iPSC-CMs derived from individuals with a disease phenotype, and that iPSC-CMs may be useful to identify individuals at risk for a cardiac event.