For the first time, a Toronto-based company called Deep Genomics has discovered a drug target and corresponding candidate using artificial intelligence (AI). The company announced yesterday that its AI-based drug discovery platform has identified a new treatment target and drug candidate for Wilson disease, a potentially life-threatening genetic disorder.
Kindersps | Shutterstock
People with Wilson disease cannot clear excess copper from the body, meaning it instead builds up in various tissues such as the liver and the brain. The condition affects about one in every 30,000 people worldwide and, if left untreated, it can cause life-threatening organ damage.
Traditional approaches have been a “gambling game”
Founder and CEO of Deep Genomics, Brendan Frey, says that making drugs has traditionally been a gambling game: “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 “throwing a stick into the tree and seeing what happens,” as he puts it, Frey built Deep Genomics, a company that uses artificial intelligence to discover new disease targets as well as the best corresponding compounds to treat them.
Frey is a professor of electrical and computer engineering at the University of Toronto. He first began building “primitive” versions of the AI model in 2004. Over the following ten years, he and his team developed the AI system and by 2014, they had created one that was better than people at interpreting human genetic mutations.
After finding that the corporate medical world did not take an interest in his work as quickly as he had expected, Frey took matters into his own hands in 2015 and set up Deep Genomics.
A new era of drug discovery
The company uses more than twenty “carefully validated and tested” machine learning systems that have been trained on public and proprietary data to screen for disease-causing mutations in the hunt for novel drug targets. Within two hours, the system can scan over 200,000 pathogenic patient mutations and automatically identify potential drug targets.
I am certain that we are witnessing a new era of drug discovery… Our AI systems can figure out how diseases are caused and how to fix those diseases much more rapidly than humans ever could and also with a greater success rate.”
Brendan Frey
The treatment target and drug candidate that has been identified for Wilson disease is the first successful drug discovery to be achieved using the platform and the first-ever example worldwide of AI being used to identify a drug candidate.
Challenges in treating Wilson disease
The current drugs for Wilson disease aim to lower copper levels by stopping the absorption of copper from food or by making the body excrete copper through urine. No treatments have yet been developed that restore the ability to remove copper because scientists have not yet understood the genetic mutation that causes the 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,” says Frey. “But there is a huge space of mutations out there that don’t cause a problem through that type of mechanism.”
Using AI to discover new targets for Wilson disease
This is where Deep Genomics’ AI system comes in. It worked out that the mutation changes an amino acid in a copper-binding protein. It also predicted that the mutation should not affect how the protein functions.
The system figured out that, instead, the mutation disrupts an instruction in the genome that tells cells how to make the protein, which results in it not being made at all.
The system identified a dozen potential new drug candidates and after testing their pharmacokinetics and tolerability in the lab, Deep Genomics announced the first pipeline candidate it intends to develop into an investigational new drug.
Frederick Askari, Director of the Wilson Disease Program at the University of Michigan, says Researchers have struggled for two decades, without success, to understand the mechanism of the genetic variant that causes Wilson disease:
The clarity that this artificial intelligence platform has brought to the scientific community is astounding and the potential of a therapy that could operate at the genomic level to correct the disease process is exciting.”
Frederick Askari
Taking drug discovery from 17 years to 18 months…
Frey says that drug discovery and development is usually a very lengthy process and that taking a patient’s disease, trying to figure out how it works and eventually reaching the stage of getting a drug approved, takes around 17 years.
By using AI, Deep Genomics has taken just 18 months to progress from identifying a new treatment target to finding a corresponding candidate drug.
“This is truly unprecedented and opens the door to a smarter, faster, and vastly more efficient means of identifying viable drug candidates for a host of diseases,” says Arthur Levin, a member of Deep Genomics strategic advisory board.
“Developing new therapeutics is full of unknowns, but I am certain that we are witnessing a new era of drug discovery.”
Frey says he anticipates that the system will get even faster at generating drug candidates and that the team expects to declare two candidates this year, at least twice as many next year and at least twice as many as that the following year.
Getting the drugs to patients
The challenge Deep Genomics will now face it how to develop all of these candidates into approved drugs that can be delivered to patients.
We [are] 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.”
Brendan Frey
Head of Johnson & Johnson Innovation JLABS in Toronto, Allan Miranda, referred to this shift in drug development as both “exciting and humbling.” He says AI will provide the opportunity to diagnose patients early and predict the disease course so that the disease can be treated early, potentially before symptoms manifest.
“Eventually, using AI in drug development will improve outcomes for patients, as better therapies reach patients sooner — which is good news in a world where ‘patients are waiting’,” he concludes.
KiNet web portal enhances study of kinase functions in cellular signaling pathways