Novoheart: Making Drugs Safer

Company Spotlight
December 15, 2017 By Wendy Wolfson

Global biotech Novoheart recently opened an R&D office at the Cove, their first U.S. location. The company is developing a drug screening platform pairing machine learning with human engineered cardiac cell and tissue assays to better predict cardiac toxicity and efficacy of drugs. Co-founded by Professors Michelle Khine, Ronald Li, and Kevin Costa, the company was listed on the Toronto Stock Exchange’s TSX Venture Exchange in October 2017.

Detecting negative side effects and drug-drug interactions, in particular cardiac toxicity, early in the drug development process is crucial to improve safety and reduce cost. Compounds with unforeseen cardiotoxicities still slip through the FDA approval process ─ indeed, the primary reason that drugs are withdrawn, either in the clinical phase or post-FDA approval, is cardiotoxicity. For example, Cisapride, a heartburn drug, was withdrawn from the market after FDA approval because it caused serious ventricular arrhythmias and sudden deaths in some of the people who took it; ultimately, the drug was linked to 300 deaths, 16,000 injuries and cost its maker $90 million in lawsuits. Use of Flecainide, an anti-arrhythmia medication, was restricted after the CAST trial was ended prematurely as the drug caused certain heart patients to die suddenly by increasing lethal arrhythmias.

Eliminating unsuitable drug candidates earlier in the pipeline could reduce the overall cost of drug development. Currently, less than 1% of initial drug candidates reach commercialization. Factoring in the attrition costs of all the other compounds that fail for every successful candidate, it takes an estimated two to four billion dollars and over 10 years to develop a new drug in the U.S.

Detecting harmful heart effects before anybody gets hurt

According to Dr. Gabriel Wong, VP for Scientific Development at Novoheart, the current FDA-recommended test for evaluating the cardiotoxicity of new drugs uses non-cardiac/non-human cell lines, such as immortalized transgenic Chinese hamster ovary (CHO) cells or human embryonic kidney (HEK293) cells, which are modified to overexpress a single protein ─ the cardiac ion channel protein hERG (human Ether-à-go-go-Related Gene)*. hERG was artificially put into CHO/HEK293 cells because scientists at the time could not readily grow or study human cardiomyocytes for drug testing. “As one can readily imagine, the current standards are inaccurate because native heart cells carry a library of proteins and channels other than just the single hERG. Animals are another widely-used alternative, but they, too, cannot accurately model the human heart,” said Wong. “As such, the miss rates (both false positive or false negative) in drug development are high, leading to the release of numerous unsafe drugs to the clinical trial pipeline, and even the market. Given the obvious need for better alternatives, the FDA is embracing newer technologies, such as our MyHeartTM Platform, that use human pluripotent stem cell-derived heart cells and engineered tissues to model the native human heart and its responses to drug compounds.”

The MyHeart drug screening platform currently includes three assays;

  • hvCAS ─ a two-dimensional, aligned, multicellular visual assay for quantifying risk of cardiac electrical disturbances
  • hvCTS ─ a three-dimensional tissue strip that models the cardiac muscle
  • hvCOC ─ a three-dimensional pumping organoid chamber that models a human ventricle

“A major difference between us and competitors is that we are not doing traditional cells in a dish-type format,” said Dr. Eugene Lee, Novoheart data scientist. “For example, we’ve been able to capture re-entrant arrhythmias in the custom-designed hvCAS assay, something that simply has not been possible with traditional 2D cultures.”

Interpreting the cell information

One of the biggest challenges is that cardiotoxicity is not just one single phenomenon ─ it comes in multiple forms,” said Lee. “Different diagnostic platforms are often difficult to compare as they vary in the information they provide because they use different tissue formats and cell sources for assays.  Meanwhile, the increasing sophistication and throughput of these cellular drug screens allow testing for an increasing number of variables and creating larger multidimensional datasets. Making sense of all these data becomes a challenge.”

Automating screening

Novoheart is now leveraging machine learning to interpret the data generated using their MyHeart platform. The company is demonstrating a computational model to predict the mechanistic action of an unknown compound, examining and classifying the drug’s overall effect on cardiac tissues. In a recent study, jointly published by Novoheart and the UC Irvine team in the journal Stem Cell Reports [], researchers used the company’s hvCTS tissue strip assay to demonstrate proof of concept for their machine learning approach, querying 17 different parameters. “The idea was to embrace the complexity of the captured data signals without trying to oversimplify it,” said Novoheart engineer David Tran, Ph.D., “and then train the computer to make sense of it all.”

How Novoheart got started 

Certain inventions that underlie Novoheart technology were developed by UCI Professor of Biomedical Engineering, Michelle Khine, in collaboration with Dr. Ronald Li, CEO of Novoheart, to design micro-fabrication technologies that would align human heart cells in a structure that would mimic native human heart cells. Novoheart licensed this intellectual property from the Regents of the University of California, and the company has engaged with Applied Innovation for an industry-sponsored research project, before becoming a tenant at The Cove in October 2017.

*The hERG gene is the human version of the Ether-à-go-go gene in the Drosophila fly; thus named in the 1960s by William D. Kaplan at the City of Hope Hospital in Duarte, California because when fruitflies with mutations in the Ether-à-go-go gene were anaesthetised with ether, their legs shook, reminiscent of the dancing then popular at the Whisky A Go-Go nightclub in West Hollywood, California.