Drug discovery needed to change. The low-hanging fruit had been harvested, but the biopharma industry, in the words of Deep Genomics CEO Brendan Frey, is still shoving the tree until an apple falls.
“Making drugs has traditionally been a gambling game. Big Pharma is throwing a stick into the tree and seeing what happens,” Frey told FierceBiotech. “It’s like the Big Pharma companies come into a casino, put a million-dollar coin into a slot machine and with some probability like 10% or something, they get a win.”
Instead of gambling to get at the fruit higher up on the tree, Frey built Deep Genomics, a company using artificial intelligence to discover new disease targets as well as the best compounds to drug them. He calls it building a ladder.
The Toronto-based biotech has been quietly plugging away since 2015 on its AI Workbench, a technology driven by—you guessed it—AI to discover genetic medicines. It uses more than 20 machine learning systems that have been “carefully validated and tested” and trained on public and proprietary data to screen disease-causing mutations in search of new drug targets.
“We have built a system that within two hours can scan over 200,000 pathogenic patient mutations and automatically identify potential drug targets,” Frey said. Instead of chasing known targets or getting into a hot therapeutic area, Deep Genomics is letting the system do the choosing—and it landed on Wilson disease, a rare genetic disorder that has no disease-modifying treatments.
People with Wilson disease can’t clear excess copper, which ends up accumulating in various tissues such as the liver and the brain. Current drugs focus on lowering copper levels by stopping the body from absorbing the copper in food or making it get rid of copper through urine. If untreated, Wilson disease is fatal.
There is no treatment that helps patients regain the ability to remove copper, because researchers struggled to understand how the underlying mutation led to Wilson disease.
“Mutations can cause disease in different ways and humans are very familiar with some of those ways, such as when a mutation changes a protein so it no longer works correctly,” Frey said. “But there is a huge space of mutations out there that don’t cause a problem through that type of mechanism.”
And that is where AI comes in. Deep Genomics’ system figured out that the mutation changes an amino acid in ATP7B, a copper-binding protein that is absent in Wilson patients. But it also predicted that this change should not affect how the protein works. The AI eventually discovered that the mutation disrupts an instruction in the genome that tells cells how to make that protein, causing it to not be made at all.
The platform identified a dozen potential drug candidates, and Deep Genomics took them to the lab to make sure they worked the way the AI predicted they would. After some tolerability and pharmacokinetics experiments, the company declared DG12P1 its first pipeline candidate that it would advance toward IND.
“We view this first declaration as ushering in a new era in drug discovery, for us, Deep Genomics, as well as more broadly,” Frey said. “Because this is really the first time that AI has really helped many different steps of drug discovery.”
Other companies have deployed this technology at various stages of discovery. Insilico Medicine recently published a study showing that in just three weeks, its algorithm turned up 30,000 new compounds to target discoidin domain receptor 1, or DDR1, which is involved in fibrotic diseases. Atomwise’s technology screens virtual compounds to identify new drugs that might take years to find using traditional discovery methods. And Insitro is using in vitro systems to predict what drug developers would see in a human clinical model. It is these models that would propose drug targets and predict how patients, or specific subgroups of patients, will respond to certain treatments.
For Deep Genomics, it took 18 months to get from a brand-new target to a drug candidate. And it may nominate new programs even faster.
“We expect to declare two candidates this year, we expect to declare at least twice as many next year and at least twice as many as that the year after that,” he said.
The challenge will be how Deep Genomics manages to develop all of these candidates and get them to patients.
“We going to do that through a combination of developing them internally and also partnering them out. It’s something we would do very carefully because we want to make sure that drugs are developed efficiently,” Frey said.
Will Deep Genomics’ approach become the ladder biopharma needs to reach further up the apple tree? Time will tell.