1 Scientific RepoRts | 5:11817 | DOi: 10.1038/srep11817 www.nature.com/scientificreports Machine learning plus optical fow: a simple and sensitive method to detect cardioactive drugs eugene K. Lee 1,* , Yosuke K. Kurokawa 2,* , Robin tu 3 , Steven C. George 2,4,* & Michelle Khine 1,5,* Current preclinical screening methods do not adequately detect cardiotoxicity. Using human induced pluripotent stem cell-derived cardiomyocytes (iPS-CMs), more physiologically relevant preclinical or patient-specifc screening to detect potential cardiotoxic efects of drug candidates may be possible. However, one of the persistent challenges for developing a high-throughput drug screening platform using iPS-CMs is the need to develop a simple and reliable method to measure key electrophysiological and contractile parameters. To address this need, we have developed a platform that combines machine learning paired with brightfeld optical fow as a simple and robust tool that can automate the detection of cardiomyocyte drug efects. Using three cardioactive drugs of diferent mechanisms, including those with primarily electrophysiological efects, we demonstrate the general applicability of this screening method to detect subtle changes in cardiomyocyte contraction. Requiring only brightfeld images of cardiomyocyte contractions, we detect changes in cardiomyocyte contraction comparable to – and even superior to – fuorescence readouts. This automated method serves as a widely applicable screening tool to characterize the efects of drugs on cardiomyocyte function. Te current state of drug development is costly and inefcient: only 1 out of 5,000 compounds available at the drug discovery stage will achieve Food & Drug Administration (FDA) approval, and the process takes approximately 14 years at a cost of 1.5 billion U.S. dollars 1 . A signifcant portion of this cost is attributed to withdrawal of drugs in clinical phases or post-FDA approval, 30% of which is from cardio- toxicity 2 . For example, the diuretic drug Cisapride’s undetected cardiotoxic efects resulted in 175 deaths and 386 cases of serious ventricular arrhythmia before it was removed from the market in 2000 3 . Clearly current preclinical screening methods do not adequately detect cardiotoxicity. Te advent of human induced pluripotent stem cell-derived cardiomyocytes (iPS-CMs) creates the possibility of a better in vitro model of the human myocardium for various applications including drug screening 4–6 . While current protocols result in iPS-CMs that are immature 7 , they do express several important phe- notypic characteristics including key contractile and channel proteins 8 . However, a persistent challenge for developing a high-throughput drug screening platform using iPS-CMs is the need for a simple and reliable method to measure key electrophysiological and contractile properties. 1 Department of Biomedical engineering, University of california, irvine, irvine, cA 92697. 2 Department of Biomedical engineering, Washington University in St. Louis, St. Louis, MO 63130. 3 Department of Statistics, california Polytechnic State University, San Luis Obispo, San Luis Obispo, cA 93410. 4 Department of energy, environment, and chemical engineering, Washington University in St. Louis, St. Louis, MO 63130. 5 Department of chemical engineering and Material Science, University of california, irvine, irvine, cA 92697. * these authors contributed equally to this work. correspondence and requests for materials should be addressed to M.K. (email: mkhine@uci.edu) received: 08 January 2015 Accepted: 27 May 2015 Published: 03 July 2015 OPEN