University of Central Florida researchers are advancing AI-assisted drug screening technology with a new method that not only improves their own model’s predictive ability but also that of seven other state-of-the-art models.

This new method can be beneficial in accelerating the development of life-saving medicines that otherwise take billions of dollars and decades of time to produce.

The results were published recently in the journal Briefings in Bioinformatics.

Their new model, BindingSite-AugmentedDTA, uses their previously reported model, AttentionsiteDTI, as the first step of a two-step prediction approach.

“A unique aspect of our approach is that it can be easily integrated with any deep learning-based prediction model, which allows for improved performance compared to using the prediction models alone,” says study co-author Ozlem Garibay ’01MS ’08PhD, an assistant professor in the Department of Industrial Engineering and Management Systems.

“By integrating our approach with other state-of-the-art deep learning-based drug-target-affinity prediction models, we have shown significant improvement in prediction performance across multiple metrics,” Garibay says. “This integration makes it a powerful tool for drug discovery research.”

How it Works

The researcher’s AttentionsiteDTI model is a classification model specifically designed to determine two key aspects. First, it identifies whether a drug compound binds with a target protein, and second, it determines the specific binding site on the protein where the drug compound interacts.

Their improved BindingSite-AugmentedDTA model follows a two-step prediction approach in which the first step uses the AttentionsiteDTI model to identify the specific binding site on the protein.

In the second step, a regression prediction model is integrated to estimate the binding strength, or affinity, between the drug molecule and the identified protein binding site.

Garibay says that this combined approach enhances the accuracy of drug target affinity predictions by reducing the search space of potential-binding sites of the protein in the first step, thus making the binding affinity prediction more efficient and accurate.

The researchers validated the prediction power of their model through in-vitro experiments and used it to successfully predict binding affinity values between FDA-approved drugs and key proteins of SARS-CoV-2.

They also showed improved performance of state-of-the-art predictive models, such as GraphDTA, DGraphtDTA and DepGS, in finding the most probable binding sites of proteins when AttentionSiteDTI was included in the models compared to when it wasn’t.

Next Steps

The researchers are working on a Python package that includes most of the drug-target interaction and drug-target affinity models and datasets, which is highly customizable.

“This will enable further high-quality research in the community by providing a convenient tool for researchers to develop and evaluate their models,” Garibay says.

They also plan to make their largest model available online for inference.

“This will facilitate fast drug screening for biology and pharmaceutical researchers with limited computer science knowledge — allowing them to easily predict drug-target binding affinities and identify potential drug candidates,” Garibay says. “This can potentially accelerate the drug discovery process and lead to the development of new treatments for various diseases.”

About the Team

Ozlem Garibay is an assistant professor of Industrial Engineering and Management Systems, part of UCF’s College of Engineering and Computer Science, where she directs the Human-Centered Artificial Intelligence Research Lab. Prior to that, she served as the director of research technology. Her areas of research are big data, social media analysis, social cybersecurity, artificial social intelligence, human-machine teams, social and economic networks, network science, STEM education analytics, higher education economic impact and engagement, artificial intelligence, evolutionary computation and complex systems. She earned her master’s and doctorate in computer science from UCF.

Mehdi Yazdan-Jahromi is a third-year doctoral student in computer science at UCF. His current research interests include computer vision, drug–target interaction and algorithmic fairness.

Niloofar Yousefi’17PhD is a postdoctoral research associate at UCF’s Complex Adaptive Systems Laboratory in the College of Engineering and Computer Science. Her research areas include machine learning, artificial intelligence and statistical learning theory to develop data analytics solutions with more transparency and explainability.

Collaborators:

Aida Tayebi is a third year doctoral student at UCF. Her current research interests include algorithmic fairness and bias mitigation techniques in DTI.

Elayaraja Kolanthai is a postdoctoral research associate at the UCF Department of Materials Science and Engineering. His current research interests include the development of nanoparticles, layer-by-layer antimicrobial/antiviral nanoparticle coatings, polymer composites for tissue engineering, and gene/drug delivery methodologies.

Craig Neal’14 ‘16MS ‘21PhD is a postdoctoral research associate at the UCF Department of Materials Science and Engineering. His current research interests include wet chemical synthesis and surface engineering of nanoparticles for biomedical applications and electrochemical devices, and electroanalysis of nanomaterials and bio-nano interactions.

Sudipta Seal is currently the chair of the Department of Materials Science and Engineering at UCF as well as a Pegasus Professor and a University Distinguished Professor. He joined the Advanced Materials Processing and Analysis Center (AMPAC) at UCF in 1997. He has been consistently productive in research, instruction and service to UCF since 1998. He has served as the nano initiative coordinator for the vice president of research and commercialization. He served as the director of AMPAC and the NanoScience Technology Center from 2009 to 2017.

Research Study: BindingSite-AugmentedDTA: enabling a next-generation pipeline for interpretable prediction models in drug repurposing