It can take, on average, more than a decade and about $1-billion for a new pharmaceutical drug to make its way from the lab to the prescription pad.
Just five in 5,000 drugs that enter preclinical testing advance to human clinical trials. From there, only about one in five of those drugs is approved for human use, according to a review by the California Biomedical Research Association.
“There are many reasons why it takes so long and costs so much money,” says Naheed Kurji, president and chief executive officer of Toronto-based Cyclica Inc., an artificial intelligence (AI)-driven biotech drug discovery platform. “When you take a drug, and you place it into a complex biological system like a human or an animal, it’s interacting with upwards of 300 proteins. And those other proteins are not known, initially. They’re oftentimes undesirable and they can lead to side effects.”
These side effects are one of the main reasons only one in 5,000 potential drugs ever makes it to a medicine cabinet.
Cyclica harnesses AI and machine learning, along with a vast library of global human genome discovery, to model potential protein interactions and drastically speed up the drug discovery process.
“We are building the biotech pipeline of the future,” Mr. Kurji says.
The seed of Cyclica was planted in 2011 at an MBA business case competition at the University of Toronto’s Rotman School of Management, presented by company co-founder Jason Mitakidis.
The proposal won the competition “hands down,” says Mr. Kurji, who was in the audience that day. Cyclica launched in 2013. Mr. Kurji joined shortly after as co-founder and chief financial officer and became president and CEO when Mr. Mitakidis left the company in 2016.
From humble beginnings in a basement office with a small team of co-op students, today Cyclica has more than 70 employees and advisers at its headquarters in Toronto, a team in the U.S. and another in the United Kingdom. The company has consultants all over the world and partnerships with biotech players in Brazil, Singapore, Korea, China, the U.S., Europe, the U.K., India and Australia, among others.
The evolving genome landscape
Disease is most often a malfunctioning of a biological protein in the human body. Computational techniques have been used for decades to pinpoint these biological drivers of disease, the malfunctioning proteins, and then find a molecular “key” that could be turned into medicine to address the malfunction. But those earlier efforts were limited.
“The techniques that they were using were too slow, they were too expensive and the quality of the predictions just were not that high,” Mr. Kurji says.
Then three things happened that drastically changed the landscape, he says: First, the Human Genome Project produced reams of data on genetics and the genome. Second, the cloud made available unprecedented computational horsepower. And third, AI and machine learning began to take hold.
A field of about 15 companies in the space when Cyclica launched has grown to more than 400 worldwide today.
How Cyclica is using AI
Cyclica has two platforms powered by the Google Cloud: Ligand Design and Ligand Express.
The underlying technology of these platforms is an AI-driven database of all publicly available known protein structures, as well as third-party proprietary data that Cyclica has acquired. Recently, the company integrated Google Deep Mind’s Alpha Fold 2 protein structure database, as well.
After pinpointing the malfunctioning protein that is the root cause of disease, the next step in drug development is to identify a molecule that will bind with that protein to address the malfunction.
Cyclica’s platforms can investigate molecules by matching them against all the proteins in the human body, explains Andreas Windemuth, the company’s chief scientific officer.
Traditionally, this research takes a target-based approach, examining the molecule for the one function it is hoped to affect.
”What our platform does is really provides a panoramic view of the molecule,” he says.
Cyclica’s database makes available approximately 85 per cent of the human proteome collection of all human proteins as well as other species.
“We’re sort of packaging all the knowledge about the drug-protein binding into our AI model and that can then be applied for discovering drugs,” Dr. Windemuth says.
The AI system keeps getting better over time as more data are added, adds Stephen MacKinnon, Cyclica’s vice-president of research and development, and it operates much faster than other forms of prediction.
”That’s what allows us to extrapolate those predictions to many, many more proteins … not just predict for that one target protein in the tunnel, but for all the proteins in the cell,” Dr. MacKinnon explains.
In short, Cyclica’s Ai-driven platforms can test thousands of proteins and millions of molecules in a fraction of the time.
Dr. Windemuth says the hope is that by speeding up and streamlining the drug discovery process, development costs will decrease and, ultimately, the cost of drugs to consumers will go down as well.
”Every month [in development] is worth many millions of dollars … and the failure rate is enormous,” he says. “We can make it faster, and we can reduce the failure rate.”
Shifting focus to ideate and create
Cyclica has switched gears from its initial focus of licensing its technology to the pharmaceutical industry. The company now sometimes partners with early-stage biotech companies working on a specific disease, becoming investors and using their technology to advance drug development, or with academic groups looking to commercialize their research.
But the primary focus is their own drug discovery pipeline.
”We recognized that to capture the value that our platform was creating, we wouldn’t do that through just revenue-generating deals with Big Pharma. We had to ideate, create and invent our own drug discovery pipeline,” Mr. Kurji says.
The company recently collaborated with researchers at the university formerly known as Ryerson, the University of Toronto and the Vector Institute to explore existing drugs that might be repurposed to treat symptoms of COVID-19. The results, which identified a drug currently used to treat lung cancer, are currently being submitted for peer review.
Over the past three years, Cyclica has created about eight companies and has more than 80 programs in its portfolio. None is in the clinical phase yet, Mr. Kurji says.
”There’s no AI and drug discovery company that has a drug that has gone through the clinical [phase] to market approval. It’s still too soon,” he says. “In a space that’s only eight years old … but there’s been a substantial amount of progress across the industry.”
How Cyclica is using AI to fight rare disease
CDKL5 Deficiency Disorder (CDD) is a rare genetic condition that affects one in every 40,000 to 60,000 children born.
A genetic form of epilepsy, CDD affects mostly girls and it can have devastating symptoms that include the onset of severe seizures as early as a week after birth.
”It is honestly devastating for the child because it stops all the developmental process,” says Cleber Trujillo, the lead senior neuroscientist at Stemonix, a subsidiary of Vyant Bio Inc., a biotech drug discovery company based in New Jersey. ”They can be really frequent, several times per day, these seizures.”
The disorder is caused by a mutation in the CDKL5, or cyclin-dependent kinase-like 5, which is the gene responsible for creating a protein necessary for normal brain development and function. The exact reason for the mutation is unknown and there is no treatment or cure.
Cyclica and Vyant Bio recently announced a strategic collaboration to use Cyclica’s AI-driven platform to identify potential pathways to the treatment of the disorder.
Vyant has exceptionally good models for the disease activity, Dr. MacKinnon says. And Cyclica has an AI-driven database of global human genome information that helps researchers such as Vyant Bio to identify and model potential target proteins that can be used to build a drug to treat the disorder.
”This really exemplifies partnership, as the researchers coming to us have a good sense of the biology, have these good models for how a disease exists in a cell and we work together to come up with drugs or drug candidates, that will likely have these effects on the systems that they’re looking to achieve for therapeutic outcomes,” Dr. MacKinnon says.
The aim is to find target molecules, Dr. Trujillo explains, and then search or screen for compounds that can interact with the target to improve the cell’s biology.
Cyclica’s biotech pipeline means researchers don’t start from scratch when looking for proteomes that could potentially work, he says.
“It’s really hard to find a drug from billions of different possibilities,” Dr. Trujillo says. “They can create a list that we think are the top candidates.
”If we can, in collaboration [with Cyclica], narrow down and join efforts on the biology side or the modelling side, with their expertise, I feel that we can accelerate and make better models and find better compounds.”
CDD is a rare disorder but one that is becoming more prevalent, owing largely to a better understanding of the disorder and better screening, he says.
The disorder significantly shortens the lives of sufferers, Dr. Trujillo says, whether from the disease itself or the severe seizures that can cause massive neurological damage.
“It’s devastating for the family and caregivers, also,” he says.