Insilico Medicine has signed a two-program artificial intelligence (AI) drug discovery collaboration agreement with Jiangsu Chia Tai Fenghai Pharmaceutical (CTFH), a company based in China.
The program will focus on previously undruggable targets and use Insilico’s next-generation AI platform. The company – which recently raised $37m in Series B funding – will be eligible to receive up to $200m based on milestone achievement and potential royalties.
Alex Zhavoronkov, PhD, founder and CEO of Insilico Medicine, told us the goal of the collaboration is to use “the most advanced artificial intelligence systems that go substantially beyond the recently published and very popular GENTRL to discover novel molecule with the desired properties that would modulate first-in-class targets in oncology and possibly beyond.”
GENTRL, or generative tensorial reinforcement learning, is a deep generative model that optimizes synthetic feasibility, novelty, and biological activity, according to the company, which recently demonstrated its ability to accelerate research.
The deal makes two points clear, Zhavoronkov said, the first of which is that the market is maturing, with more researchers “exploring new ways to make R&D faster, cheaper” – willing to “try something that is not possible using other methods.”
It also demonstrates that the generative chemistry technology is becoming more established, he said, noting that when Insilico first presented its work in 2016 and 2017, “everyone was very skeptical.”
“I remember when I presented our paper at the first "Advanced Pharma Analytics" conference in Basel in January 2017 a lot of people said that I should not even be going this way and to make a convincing story I need to show the structures that were tested in enzymatic assays,” he said.
“Now that we went much further, people are rightfully asking for human data,” he added. “Unfortunately, the time it takes to develop and train a new generation of the AI system is 1/100s of the time it takes to validate in animals and get to humans. In other areas of AI, it is the other way around. You spend 99% developing and training and 1% of the time testing.”