Developing life-saving drugs could take billions of dollars and decades, but researchers at the University of Central Florida aim to speed that process up with a new AI-based drug screening process they’ve developed.
Using a method that models drug-target-protein interactions using natural language processing techniques, the researchers achieved an accuracy of 97% in identifying promising candidate drugs. The results were recently published in the journal Briefings in Bioinformatics.
This technique represents interactions between drugs and proteins through words for each protein-binding site and uses deep learning to extract features that govern the complex interactions between the two.
Ozlem Garibai, co-author of the study, an assistant professor at the University of California at Department of Industrial Engineering and Management Systems. “You can try many different forms of proteins and drug interactions and see which one is most likely to bind or not.”
The model they developed, known as AttentionSiteDTI, is the first to be interpreted using the language of protein binding sites.
This work is important because it will help drug designers identify important protein binding sites along with their functional properties, which is key to determining whether a drug will be effective.
The researchers achieved the breakthrough by devising a self-attention mechanism that makes the model recognize which parts of a protein interact with drug compounds, while achieving advanced prediction performance.
The mechanism’s self-attention ability works by selectively focusing on the most relevant parts of the protein.
The researchers validated their model using laboratory experiments that measure binding interactions between compounds and proteins and then compared the results with those that their model predicted computationally. As drugs used to treat COVID remain of interest, trials have also included testing and validating drug compounds that may bind to the spike protein of SARS-CoV2.
The high concordance between laboratory results and computational predictions demonstrates the potential of AttentionSiteDTI to pre-screen potentially effective drug compounds and accelerate the exploration of new drugs and the reuse of existing drugs, Garibai says.
Sudipta Cel, co-author of the study and president of the University of California at F Department of Materials Science and Engineering.
Mehdi Yazdani Jahromi, PhD student at the University of California College of Engineering and Computer Science The work offers a new direction in drug prescreening, says the study’s lead author.
“This allows researchers to use AI to more accurately identify drugs to respond quickly to new diseases,” Yazdani Jahromi says. “This method also allows researchers to determine the best binding site for the virus protein to focus on in drug design.”
“The next step in our research will be to design new drugs using the power of artificial intelligence,” he says. “This could naturally be the next step in preparing for a pandemic.”
The research was funded by the Initial Funding Program for Big Data and Internal Artificial Intelligence at UCF.
Study co-authors also included Nilufer Yousefi, a postdoctoral researcher at UCF Complex adaptive systems lab in the School of Engineering and Computer Science at UCF; Aida El-Teibi, a doctoral student in the Department of Industrial Engineering and Management Systems at the University of California. Elayaraja Kolanthai, is a postdoctoral research associate in the University of California’s Department of Materials Science and Engineering. and Craig Neal, a postdoctoral researcher in the University of California’s Department of Materials Science and Engineering.
Garibai received her Ph.D. in Computer Science from the University of California, and joined the UCSD Department of Industrial Engineering and Management Systems, part of the College of Engineering and Computer Science, in 2020. She previously worked for 16 years in Information Technology at UCLA. Research office.
Article: AttentionSiteDTI: an interpretable graph-based model for predicting drug-target interaction using sentence-level relationship classification in NLP
Contact: Robert H. Wells, Office of Research, 407-823-0861, [email protected]
Link to the magazine article: https://academic.oup.com/bib/article/23/4/bbac272/6640006