DALLAS, October 7, 2024 -- Drug discovery is much like working a jigsaw puzzle. The chemical compounds behind drug molecules must be shaped to fit with the proteins in our bodies to produce therapeutic effects. That requirement for a meticulous fit means the creation of new drugs is extremely complex and time-consuming.

To speed up the puzzle-fitting process, researchers at SMU have created SmartCADD. This open-source virtual tool combines artificial intelligence, quantum mechanics and Computer Assisted Drug Design (CADD) techniques to speed up the screening of chemical compounds, significantly reducing drug discovery timelines. In a recent study published in the Journal of Chemical Information and Modeling, researchers demonstrated SmartCADD’s ability to identify promising HIV drug candidates.

“There is an urgency to discover new classes of drugs like antibiotics, cancer treatments, antivirals and more,” said Elfi Kraka, head of the Computational And Theoretical Chemistry Group (CATCO) at SMU. “Despite AI’s rapid adoption in many fields, there has been a hesitancy for using it in scientific research, mainly because of its opaqueness and the quality of data used for training. SmartCADD addresses those concerns and can sift through billions of chemical compounds in one day, which significantly reduces the time needed to identify promising drug candidates.”

SmartCADD combines deep learning models, filtering processes and explainable AI to screen databases of chemical compounds that are used to pinpoint drug leads. The tool has two main components: SmartCADD’s Pipeline Interface, which collects data and runs filters, and its Filter Interface, which tells the system how each filter should work. These built-in filters assist with different stages of testing chemical compounds. They can help predict how a drug will behave in the body, model what drug structures will look like using 2D and 3D parameters and utilize an AI model that explains its decisions.

Researchers demonstrated the SmartCADD platform through three different case studies of drugs used to treat HIV, finding that several proteins exist within the virus are believed to be promising targets. SmartCADD used data from the MoleculeNet library to create and search through a database of 800 million chemical compounds and determined that 10 million might work as HIV drugs. It then used filters to find compounds that best matched already approved HIV drugs.

While the researchers focused on HIV targets for the study, they stressed that SmartCADD is versatile and can be applied to other drug discovery pipelines.

“This is a user-friendly virtual screening platform that provides researchers with a highly integrated and flexible framework for building drug discovery pipelines,” said Corey Clark, assistant professor of computer science in the Lyle School of Engineering and deputy director for Research at SMU Guildhall. “We are going to continue pushing the work forward to expand chemistry and machine learning capabilities even further. The project and its opportunities are truly exciting, and I know the next phase will be an even bigger step forward than the last.”

Adds Laird, “Interdisciplinary research is absolutely necessary to make major research advancements that actually impact the real world. This is a major focus of SMU and a key reason I wanted to pursue my Ph.D. here. Impactful research can't happen in the vacuum of a single field. You have to look broadly across disciplines to spark ideas that will turn into true innovations. Breakthroughs often occur at the intersection of different fields, and that's where I aim to position my research.”