AI is used to design a drug in 21 days by start-up

AI has taken the development of drugs from the usual 8 years or so and millions of dollars in investment down to 21 days and approximately $150,000 in cost.

This has been made possible by a young company called Insilico Medicine. According to Wikipedia, Insilico is headquartered in Rockville, MD as part of the Johns Hopkins University’s Emerging Technology Centers. Insilico combines genomics, big data analysis, and deep machine learning for its drug discoveries.

Insilico was founded by CEO Alex Zhavoronkov in 2014. He says that his company has taken AI strategy and combined it with AI imagination to develop a system called generative tensorial reinforcement learning (GENTRL). Insilico published its research on Monday, Sept. 2, 2019 that showed GENTRL identified a potential drug treatment for fibrosis in just 21 days. The research was published in Nature Biotechnology,

 Zhavoronkov’s background was originally in computer science and worked for ATI Technologies Inc. (ATI) and later for Advanced Micro Devices (AMD). In 2006, Zhavoronkov switched to biotechnology research and earned his Masters degree from John Hopkins University in Baltimore, Maryland, USA, then received his PhD from Moscow State University. His studies focused on the use of ‘machine learning’ to observe the physics of the molecular interactions in biological systems. He returned to Baltimore in 2014 and founded Insilico.

Typically, in the traditional drug discovery world, screening molecules from large vendor libraries is the way they do it which takes years. Zhavoronkov and his team wanted to know if it could be done with machine learning and if it could be done more quickly.

This original idea was researched and results published in 2016 brought in investment dollars to the valuation of $56 million.

The current published study is a result of a challenge posed to Insilco by colleagues in the chemistry world to develop potential drugs that could inhibit 

discoidin domain receptor 1 (DDR1) activityin fibrosis with the use of GENTRL. This challenge was made on the recently published research by Genentech which took 8 years to identify possible promising DDR1 kinase inhibitors through traditional methods.

GENTRL designed six new drug candidates which were then synthesized and out of the group one was successfully tested in mice. It took the AI system 21 days to design the molecules and the total time from design, synthesis and validation took about 46 days.

Although all of the drugs for DDR1 drug inhibitor designed by Insilco’s GENTRL do not appear to be more effective than those produced by the traditional research method, GENTRL definitely is a process that saves in time and money and shows the promise of using AI for drug discovery.