Last year, out of thousands of drugs in clinical trials, the FDA approved only 37, the fewest to pass regulatory scrutiny since 2016. Now, artificial intelligence could soon help figure out which drugs will work and which ones will not.
[Ads /]
"We can cut down the experiments as well as cut down the time," said Dr. Sudipta Seal, a professor in the department of Materials Science and Engineering at the University of Central Florida.
On average, it takes 10 years to develop a new drug. Now, computer scientists are aiming to speed up that process using an AI-based drug screening method that does the intensely difficult job of matching drug and protein interactions.
It translates the complex interactions at each drug protein-binding site into words. The AI model then analyzes that language to learn which part of a virus protein a drug will bind to. With 97% accuracy, it can predict how well a drug will work.
"You can just, give it, for example, COVID protein and test it against all the FDA-approved drugs and see whether or not they bind or not," said Mehdi Yazdani-Jahromi, a UCF PhD student. "That's the beauty of this work."
It's called AttentionSiteDTI and the model is ready to be used now for anyone developing a new drug at no cost.
[Ads /]
"I think it's going to revolutionize medical fields in so many different ways," said UCF computer scientist Dr. Ozlem Garibay.
It's simple: the less time in the lab, the lower the cost will be to create it. Researchers believe that savings could be handed down to the consumer.
They're now ready to create a website like ChatGPT, making it simple for other scientists to put in their data and see if their drug will work or not.