It could take years for the pharmaceutical industry to develop medicines that can treat or cure human ailments, but new research suggests the use of generative artificial intelligence could greatly accelerate the drug development process. is suggested.
Today, most drug discovery is done by human chemists who rely on their knowledge and experience to select the appropriate molecules necessary to become the safe and efficient medicines we depend on. Synthesize. To identify synthetic routes, scientists often use a technique called retrosynthesis. This is a method of creating potential drugs by working backwards from the molecules of interest and exploring the chemical reactions to manufacture them.
But sifting through millions of potential chemical reactions can be a very difficult and time-consuming task, so researchers at Ohio State University created an AI framework called G.2Retro automatically generates reactions for any molecule. New research shows that, compared to current manual planning methods, this framework covers a huge range of possible chemical reactions, allowing us to accurately and precisely determine which reactions are the best to generate a particular drug molecule. shown to be readily identifiable.
Using AI in things that are critical to saving lives, such as healthcare, is something we really want to focus on. Our aim is to use AI to accelerate the drug design process, which not only saves researchers time and money, but also has properties far superior to any molecule found in nature. It has been found to provide drug candidates that may have ”
Xia Ning, lead author of the study and associate professor of computer science and engineering at The Ohio State University
The work builds on Ning’s previous work, in which her team developed a method called Modof that can generate molecular structures that exhibit desirable properties over any existing molecule. “The question becomes how to make such a product molecule, and that’s where this new research really shines,” said Ning, who is also an associate professor of biomedical informatics at the School of Medicine. Told.
The study was published today in a journal communication chemistry.
Ning team trained G2A retro dataset containing 40,000 chemical reactions collected between 1976 and 2016. This framework “learns” from a graph-based representation of a given molecule and uses deep neural networks to generate structures for potential reactants that can be used in their synthesis. Its generative power is so impressive that, according to Ning, once the molecule is given, G2Retro was able to come up with hundreds of new reaction predictions in just minutes.
“Our generative AI method G2“Retro can provide multiple different synthetic routes and options, as well as a way to rank the different options for each molecule. This is not a replacement for current lab-based experiments, but more. It will provide a better drug for the world,” Ning said. Options help you prioritize your experiments and focus more quickly. ”
To further test the effectiveness of AI, Ning’s team conducted a case study to test G2Retro was able to accurately predict four newly launched drugs already in circulation. Tapinaloff is used to treat various skin disorders. Mabacamten, a drug used to treat systemic heart failure. Oteseconazole is used to treat fungal infections in women. G.2Ning said Retro was able to produce exactly the same patented synthetic route for these drugs and also provided a viable and synthetically viable alternative synthetic route.
Putting such dynamic and effective devices at the scientists’ disposal could enable industry to produce more potent drugs at a faster pace. But despite the potential for cutting-edge AI to scientists in the lab, Ning stresses the importance of pharmaceuticals.2Retro or any generative AI artifacts still need to be validated. That process involves testing the created molecule in an animal model, followed by human trials.
“We are very excited about generative AI for medicine and are committed to using AI responsibly to improve human health,” said Ning.
This research was supported by the Ohio Presidential Research Excellence Program and the National Science Foundation. Other co-authors from Ohio State University were Ziqi Chen, Oluwatosin Ayinde, James Fuchs, and Huan Sun.
sauce:
Reference magazines:
Chen, Z., other. (2023) G2Retro as a two-step graph generation model for retrosynthetic prediction. communication chemistry. doi.org/10.1038/s42004-023-00897-3.
