Hong Kong-based Insilico Medicine published research Monday showing that its deep learning system could identify potential treatments for fibrosis. That system, called generative tensorial reinforcement learning, or GENTRL for short, was able to find six promising treatments in just 21 days, one of which showed promising results in an experiment involving mice. The research has been published in Nature Biotechnology, and the code for the model has been made available on Github.
“We’ve got AI strategy combined with AI imagination,” says Insilico CEO Alex Zhavoronkov, who compares the operation of GENTRL to the AlphaGo machine learning system that Google’s Deepmind developed to challenge champion Go players.
Zhavoronkov founded the company in 2014. His original background was in computer science, and he spent several years working at ATI until it was acquired by AMD in 2006. At that point, he shifted gears and decided to go into biotechnology research, with an interest in research into slowing down the aging process. He received his Masters from Johns Hopkins and then got a PhD from Moscow State University, where his studies focused on using machine learning to look at the physics of molecular interactions in biological systems. He then worked for a number of companies, but then returned to Baltimore to found Insilico.
The company's original philosophy was about using deep learning to train neural networks to go through large libraries of molecules to find drug targets. Shortly after founding the company, however, Zhavoronkov became fascinated with Ian Goodfellow’s work in machine learning, and decided to switch course.
The initial research the company published based around this idea in 2016 helped bring in investment money in the competitive biotech and AI arenas. According to Pitchbook, it’s so far raised $24.3 million in investment dollars at a valuation of $56 million, from backers including A-Level Capital and Juvenescence. It also has multiple partners across the biotechnology field, including A2A Pharmaceuticals and TARA Biosystems.
The current paper has its origin in a challenge put to the company by its colleagues in the chemistry world. They ask the company to use its system to develop potential drugs that can inhibit discoidin domain receptor 1 (DDR1) activity. DDR1 is an enzyme that is involved in fibrosis, and though it’s not yet clear if it regulates those processes, inhibiting its activity is being investigated as a possible therapy. The challenge was based on recently published research from a team at Genentech, which had taken about 8 years to identify promising DDR1 kinase inhibitors.
An overview of Insilico's discovery process
INSILICO MEDICINE
Insilico used GENTRL to design new drug candidates, which were then synthesized and a leading candidate was successfully tested in mice. It took about 21 days for the AI system to design molecules, and the total time for design, synthesis and validation was about 46 days. Although none of the drugs designed by GENTRL would appear to be more effective than inhibitors developed by the traditional research method, the traditional process to develop drug candidates took over 8 years and millions of dollars to develop - compared to the handful of weeks and approximate $150,000 cost of Insilico’s method.
Source:forbes.com
“Their molecules are amazing, they’re a little bit better than what our AI could do,” Zhavoronkov says. “But again that’s years versus people who don’t have a lot of knowledge of chemistry doing this stuff.”
Though he cautions that Insilico still has a lot of work ahead of it, for Zhavoronkov, this research is an important breakthrough because it shows the promise of AI for drug discovery.
“I think that this paper will remove a lot of skepticism in big pharma,” he says.
“We thought, ‘Can we make machines imagine new molecules with particular properties instead of screening large vendor libraries?’” he says. Screening molecules, he says, is how they do it in the traditional drug discovery world, but he wanted to see if this type of machine learning could do things more quickly.
No comments:
Post a Comment