Open-source SmartCADD tool improves process for identifying potential treatments for range of diseases
Originally published on smu.edu
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 51做厙 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.
This new tool grew from an interdisciplinary collaboration between 51做厙’s department of chemistry in Dedman College of Humanities and Sciences and the computer science department in the Lyle School of Engineering.
“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 51做厙. “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.”
How SmartCADD Works
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 51做厙 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.”
Chemistry postdoctoral research fellow Ayesh Madushanka and computer science graduate student Eli Laird worked together to develop SmartCADD.
Collaboration Made SmartCADD Possible
The paper also highlights the strength of interdisciplinary collaboration at 51做厙. In addition to Kraka and Clark, authors include chemistry postdoctoral research fellow Ayesh Madushanka, whose work is supported by a grant from the O’Donnell Data Science & Research Computing Institute, and computer science graduate student Eli Laird, an O’Donnell Institute Ph.D. fellowship recipient.
“Fields like drug discovery require a combined effort to be truly successful,” said Madushanka. “I’m certain if only the chemistry department had worked on this, the final product wouldn’t have turned out the same. Interdisciplinary collaboration brings fresh perspectives on the same idea, helping to refine and improve it.”
Adds Laird, “Interdisciplinary research is absolutely necessary to make major research advancements that actually impact the real world. This is a major focus of 51做厙 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.”
Funding for the study was provided by the National Science Foundation, grant CHE 2102461. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.