Researchers at the University of Oregon have developed an artificial intelligence tool designed to predict how new drug molecules move within the body before costly laboratory tests are conducted. Led by doctoral student Revanth Elangovan and postdoctoral researcher Sompriya Chatterjee under the guidance of biophysicist Dhiman Ray, the team published their findings in the Proceedings of the National Academy of Sciences. The algorithm efficiently simulates the behavior of never-before-seen molecules based on their chemical structure, offering a significant leap forward in drug discovery. While existing computational tools like Google's AlphaFold have revolutionized the field by predicting the static structure of molecules—a contribution recognized by the 2024 Nobel Prize in Chemistry—they are limited. According to Ray, these programs are akin to taking a still photograph of a movie; they capture a moment in time but fail to show the action. To determine whether a drug will bind to its intended target or interact with off-target elements, scientists must understand molecular motion and interaction over time. Traditional simulation methods that capture this movement are computationally intensive, resembling a full-length film that is too expensive and time-consuming for many labs to produce. The new approach developed by the Oregon team acts as a middle ground, comparable to a Wikipedia plot summary. It bridges the gap between a static image and a detailed simulation, providing the essential information needed to predict molecular movement in the most efficient manner possible. The model integrates artificial intelligence with physics data, using measurements of known molecule behaviors and the energy required for shape changes to guide the algorithm. This hybrid approach prevents the AI from wasting energy exploring unlikely scenarios while using mathematical techniques to increase the probability of capturing critical moments, such as the fraction of a second when a drug binds to its target. Ray envisions that within 10 to 20 years, such technology could enable real-time simulations, allowing researchers to observe how different drugs bind to proteins and select the most viable candidates for human trials. The code developed by the team is freely available for use by the broader scientific community. While the primary focus is on drug development, the methodology holds potential applications across biology and chemistry for analyzing new materials. The researchers are currently working on making the model's data more interpretable, aiming to translate the output into user-friendly visualizations that blend physics insights with machine learning advantages. This combination of fields represents the unique strength of the lab's ongoing research.