According to a recent report published in the journal Chemical Science, researchers from the University of Toronto and Insilico Medicine have utilized an AI drug discovery platform called Pharma.AI to generate a potential treatment for hepatocellular carcinoma (HCC), the most prevalent form of liver cancer.

Within just one month, artificial intelligence was able to discover a new treatment pathway for cancer by utilizing AlphaFold, an AI-based protein structure database.

The research team then produced a “novel hit molecule” that could bind to the newly discovered target without any additional assistance. Furthermore, the AI system has demonstrated the ability to predict a patient’s survival rate.

Within a mere 30 days, the research team identified the target and synthesized only seven compounds to create the potential drug.

Following a subsequent round of compound generation, the researchers came across a more powerful hit molecule. Nonetheless, any potential drug derived from it would have to undergo clinical trials before it could be employed on a large scale.

Insilico Medicine’s CEO, Alex Zhavoronkov, stated that their generative AI algorithms were able to design potent inhibitors for a target with an AlphaFold-derived structure, while the world was captivated by advancements in generative AI in art and language. The traditional method of trial and error in drug discovery is slow, expensive, and limiting, which is why AI is rapidly transforming the field.

According to Nobel Prize winner in chemistry, Michael Levitt, this research is further proof of AI’s potential to enhance the drug discovery process with its speed, efficiency, and accuracy.

By combining the predictive power of AlphaFold with Insilico Medicine’s Pharma.AI platform for target and drug design, a new era of AI-powered drug discovery may be on the horizon.

In the year 2022, AlphaFold accomplished a significant breakthrough in the fields of both AI and structural biology by accurately predicting the protein structure for the entire human genome.

Co-author and co-CEO of Insilico Medicine, Feng Ren, stated that AlphaFold’s achievement marked a new scientific frontier in predicting the structure of all human body proteins.

At Insilico Medicine, they recognized this feat as an incredible opportunity to leverage these structures within their end-to-end AI platform, to develop innovative therapies for addressing diseases with significant unmet needs. This study represents an important initial stride in that direction.

Researchers also explained how different AI information can revolutionize health care.

“What this paper demonstrates is that for health care, AI developments are more than the sum of their parts,” Alan Aspuru-Guzik, a professor of chemistry and computer science at U of T’s Faculty of Arts & Science, said. “If one uses a generative model targeting an AI-derived protein, one can substantially expand the range of diseases that we can target. If one adds self-driving labs to the mix, we will be in uncharted territory. Stay tuned!” 

A separate study published in the journal JAMA Network Open showed an AI system invented by scientists at the University of British Columbia and BC Cancer was able to predict cancer patient survival rates using doctors’ notes.

This model employs natural language processing (NLP), a facet of AI that can comprehend intricate human language.

Through NLP, the system can scrutinize physicians’ notes following an initial consultation, and identify distinct traits that pertain to each patient. The model achieved a success rate of over 80% in predicting six-month, 36-month, and 60-month survival rates.

Additionally, it can determine rates for all types of cancer, a marked improvement over prior models that were only capable of applying to certain types of cancer.

According to lead author Dr. John-Jose Nunez, a psychiatrist and clinical research fellow with the UBC Mood Disorders Centre and BC Cancer, the AI model reads the consultation document in a similar manner to a human being. The consultation document contains various patient details, such as age, cancer type, underlying health conditions, past substance use, and family histories, and the AI integrates all of this information to form a more comprehensive understanding of patient outcomes.

Traditional cancer survival rates are generally calculated retrospectively and categorized according to a few generic factors, such as cancer site and tissue type. This model was tested using data from 47,625 patients located across six BC cancer sites in British Columbia.

“Because the model is trained on BC data, that makes it a potentially powerful tool for predicting cancer survival here in the province,” Nunez said. 

“The great thing about neural NLP models is that they are highly scalable, portable and don’t require structured data sets,” he added. “We can quickly train these models using local data to improve performance in a new region. I would suspect that these models provide a good foundation anywhere in the world where patients are able to see an oncologist.”

AI could be a cutting-edge technology for future cancer care that could be applied in cancer clinics around the world.

“Predicting cancer survival is an important factor that can be used to improve cancer care,” Nunez said. “It might suggest health providers make an earlier referral to support services or offer a more aggressive treatment option upfront. Our hope is that a tool like this could be used to personalize and optimize the care a patient receives right away, giving them the best outcome possible.”