China Epicenter in the AI Drug Discovery Earthquake—Did the Plates Shift?

Sep 4, 2019 | AI, Drug Development, Drug Discovery, Machine Learning

China Epicenter in the AI Drug Discovery Earthquake—Did the Plates Shift?

Insilico Medicine recently described the revolutionary forces of AI at work—radically accelerating the drug development process—one that today takes 10 to 20 years and from $500 million to $2.6 billion, to figures far more appealing for drug companies, investors, and patients. It evidenced AI’s impact today in a study revealing a new AI system that successfully identified six substances that block a certain enzyme responsible for fibrosis and other illness in just three weeks. China may be at the epicenter of this epic force.

Study Author: Insilico Medicine

The lead author of a study published in Nature Biotechnology, Insilico Medicine was founded in 2014 as an artificial intelligence (AI) company headquartered in Rockville, MD, and Hong Kong by Alex Zhavoronkov. They were founded to pioneer applications of the generative adversarial networks (GANs), reinforcement learning, transfer learning and meta-learning for generations of novel molecular structures for the diseases with known and unknown targets, and, unlike the other competitors in the field, is focused on the development of end-to-end drug pipeline covering every step of drug discovery, clinical trials analysis, and digital medicine.

According to website CrunchBase, they have raised $14.3 million and employ approximately 50 to 75 employees. They report that they will soon announce deals with prominent biopharmaceutical sponsors.

Study Summary

As reported in the South China Morning Post (SCMP), Insilico Medicine along with Chinese CRO Wuxi and researchers from the University of Toronto published the results of their AI study in Nature Biotechnology, showcasing the power of their machine-driven “generative adversarial networks (GANs) and reinforcement learning” to drive de novo drug design. The authors proved in the study that the AI technology could help drug developers identify a large number of compounds aiming for a protein expressed in epithelial cells and associated with fibrosis– known as Discoidin domain receptor 1 (DDR1) in just three weeks—a period of time for such an activity heretofore not feasible.

Chinese Epicenter

Insilico Medicine moved to Hong Kong for a reason. Founder Alex Zhavoronkov reported to the SCMP that “Right now, China is experiencing a ‘Cambrian explosion’ in biotechnology. Many top scientists from pharma and academia moved back to China and started working on known targets.” Insilico is banking on a future where, in China, the venture can cut at least a year out of the R&D process—saving millions along the way.

China has identified AI as a strategic national industry.  The SCMP reports the technology, for example, is being used to streamline its overburdened health care system—facing imminent crisis due to its rapidly aging population, acute shortage of qualified doctors, severe regional imbalances, etc.

SCMP reports that the Chinese health-care big data industry is estimated to be over $11 billion by 2020—hundreds-if not thousands-of Chinese-based AI ventures have emerged.

Big technology players such as Tencent Holdings and Alibaba are doing major healthcare-related deals. For example, Tencent (maker of WeChat) has partnered with over 100 tier one hospitals to research AI health-related applications and just last year introduced an AI-based open platform for medical diagnostics; while Alibaba does similar deals and Baidu (China’s Google) develops open-source AI to help pathologists identify breast cancer.

Complicated Learning Curve & Complexities Capture Reality

There is always another view to be taken into consideration, and more often than not, if something sounds too good to be true….well, it could be. SCMP interviewed an AI ethics expert, Danit Gal, who noted “AI’s complicated learning curve.” Gal noted the power of the theory behind AI-driven discovery and development—simulating experiments and approximating results that support the transformation of R&D over time. Gal notes, “In practice, however, we need to account for a multitude of confounding factors and feed a considerable sample of results to increase accuracy. Errors can be fatal, especially if results are not thoroughly and continually audited.”


There are huge changes coming with AI—we can see it in many facets of life today. But we are also cognizant of the conservative and highly regulated nature of healthcare and life sciences around the world—factoring China’s boom in as well. In many ways, the market was educated about AI thanks to the huge sums of money spent by IBM to showcase Watson in a variety of industries. As we have reported, that has led to mixed results in the real world. AI is thought of as an investment in the drug discovery and development world—it will over time we think, yield big returns—but don’t expect it fast. We are still at the earliest of stages of what is an unfolding, complex, and dynamic story. With many twists and turns, cul-de-sacs and some roads that shouldn’t be traveled—we don’t see a fast, linear path to a new and transformed drug discovery and development process—but rather an incremental movement driven by bursts of energy representing dynamics such as regulatory transformation, political change, market demands, process upheaval, technological advancement, etc. An organization’s vision, leadership, strategies processes and systems—not to mention its talent—will need to evolve to better understand, harnesses, and leverage these machine-learning methods in seamless, integrated real-world processes to ensure productive, safe, and high-quality outcomes.


Pin It on Pinterest