Researchers from the University of California (UC), San Diego, have developed a novel AI platform that can generate individual drug compounds capable of inhibiting multiple molecular targets at once. They used it to synthesize 32 new drug candidates for cancer.
The POLYGON platform, short for POLYpharmacology Generative Optimization Network, uses a deep machine learning model based on generative AI and reinforcement learning.
It works by embedding chemical space, in this case, target proteins, and repeatedly sampling them to generate new molecular structures to inhibit those proteins each time. The process simulates the time-consuming chemistry involved in the earliest phases of traditional drug discovery, substantially streamlining the process.
The output is optimized by reinforcement learning, a powerful machine learning strategy in which the model is repeatedly trained by rewarding desired outputs such as predicted ability to inhibit protein targets, drug-likeness, and ease-of-synthesis and by punishing undesired outputs.
What makes POLYGON unique among AI tools for drug discovery is that it can identify molecules with multiple targets to create polypharmcology drugs, whereas existing drug discovery protocols currently prioritize single-target therapies.
These polypharmacology drugs have many applications but are difficult to design and have therefore typically been discovered by chance rather than systematically. They have potential advantages over combination therapy, including superior pharmacokinetic and safety profiles, lower likelihood of acquired resistance, and simplified therapy formulation leading to increased patient compliance.
The field of polypharmacology is still in its infancy but recent studies have begun to show how it could be useful for treating cancer. Indeed, several groups have shown that KRAS mutant non-small-cell lung cancers, typically resistant to treatment with classical single-target agents, can be effectively treated using polypharmacological compounds.
“It takes many years and millions of dollars to find and develop a new drug, especially if we’re talking about one with multiple targets,” said senior author Trey Ideker, PhD, professor in the department of medicine at UC San Diego School of Medicine and adjunct professor of bioengineering and computer science at the UC San Diego Jacobs School of Engineering. “The rare few multi-target drugs we do have were discovered largely by chance, but this new technology could help take chance out of the equation and kickstart a new generation of precision medicine.”
Ideker and team trained and verified POLYGON on a database of over a million known bioactive molecules containing detailed information about their chemical properties and known interactions with protein targets.
They used the platform to generate novel polypharmacology compounds against ten pairs of protein targets, including serine/threonine kinases, tyrosine kinases, DNA binding factors, and histone modifiers, that were known to be codependent, or synthetically lethal, in human cancer cell lines. This means that inhibiting both together is enough to kill cancer cells even if inhibiting one alone is not.
The group then employed more traditional biophysical computer simulations known as docking experiments to see how the top structures bound to their two targets. They showed that the binding had similar 3D orientations to those seen with standard single-protein inhibitors.
Finally, Ideker and co-investigators synthesized 32 molecules that had the strongest predicted interactions with the MEK1 and mTOR proteins, a pair of cellular signaling proteins that are a promising target for cancer combination therapy.
They report in Nature Communications that the drugs they synthesized had significant activity against MEK1 and mTOR in vitro, with most yielding more than a 50% reduction in each protein activity and in cell viability when dosed at 1–10 μM. There were also minimal off-target reactions with other proteins.
This suggests that one or more of the drugs identified by POLYGON could target both proteins as a cancer treatment, providing a list of choices for fine-tuning by human chemists.
“Once you have the candidate drugs, you still need to do all the other chemistry it takes to refine those options into a single, effective treatment,” said Ideker. “We can’t and shouldn’t try to eliminate human expertise from the drug discovery pipeline, but what we can do is shorten a few steps of the process.”
Despite this caution, the researchers are optimistic that the possibilities of AI for drug discovery are only just being explored.
“Seeing how this concept plays out over the next decade, both in academia and in the private sector, is going to be very exciting,” said Ideker. “The possibilities are virtually endless.”
