Insilico preps first candidate for IND studies, hoping to launch trial in IPF by end of 2021
Over the last several years, Alex Zhavoronkov has turned Insilico into one of the most well-connected AI biotechs out there. Now, the Hong Kong-based company is one step closer to reaching the clinic.
Insilico is bringing its first candidate into IND-enabling studies, Zhavoronkov announced Wednesday, with the goal of launching an in-human trial at some point later this year. The move marks what he says is a first in AI drug discovery, as both the target and small-molecule inhibitor involved in the program are completely new and AI-generated.
Zhavoronkov, who told Endpoints News he builds self-driving robots in his spare time, isn’t ready just yet to say what the target or experimental drug are, but he noted that the focus will be idiopathic pulmonary fibrosis. The target itself plays a key role in fibrotic and inflammatory pathways, sitting on the intersection of profibrotic pathways, he said.
More important to Insilico is the process by which they discovered the compound — despite using more than 60 different approaches, the main method was through its proprietary AI platform’s neural networks. Insilico researchers started by looking at how different tissues susceptible to fibrosis can change over the course of patients’ lives, but that operation is too difficult for humans to do on their own, as the requisite datasets are massive.
“We train on age, after that we re-train on fibrosis, get the first list of prioritized likely targets, and then we put them into biological pathways,” Zhavoronkov told Endpoints.
After that, Insilico applied a separate neural network to comb through databases over where governments were issuing grant research and companies were taking clinical trials. That helped Zhavoronkov determine whether the target is actually new or if other companies had already tried to make something work with it.
Lastly, Zhavoronkov wanted to be able to trust the answers his neural networks were spitting out. So he essentially asked his AI to find potential hidden links between the data to increase their confidence that a drug would work, while also looking at how certain key opinion leaders viewed the biology they were trying to crack.
Insilico developed its platform in collaboration with Big Pharmas to incorporate how the “best of the best” human researchers think about target selection, Zhavoronkov said, adding that in lots of cases, however, targets are chosen for business reasons rather than science. He wants to try to change that with his platform and Wednesday’s announcement.
“We need to find this ultimate balance because you realize that more than one target can be implicated in this disease,” Zhavoronkov said. “So it is possible there is not just one magic bullet, there could be several bullets.”
When this new program ultimately reaches the clinic, by the end of this year at the earliest, it will be a typical Phase I study. But Zhavoronkov said he has the “luxury” of experimenting with every step of the trial given other pharma companies haven’t yet entered this space in this fashion.
The immediate next steps are to continue pushing out the platform to other players. Insilico has made the core of its AI open source in the hopes of bringing down drug discovery costs across the industry. Zhavoronkov said validating this target in IPF cost Insilico only $1.8 million.
“If I do not invent, I do not live,” Zhavoronkov said. “We will never stop and we also have a huge pipeline of projects in AI.”
