Advancing Food Science and Antimicrobial Discovery with HPC
Researcher Highlight: Ilias Tagkopoulos, PhD
Ilias Tagkopoulos, PhD use HPC@UCD by combining generative AI with large-scale molecular dynamics (MD) simulations. His work focused on understanding how new antimicrobial peptides—molecules could help combat antibiotic-resistant infections—an essential step toward developing novel therapeutics.
Ilias Tagkopoulos, PhD and Cheng-En Tan, PhD candidate developed a generative AI framework to design novel antimicrobial peptide (AMP) sequences, which are then tested using large-scale molecular simulations.
These simulations generate large volumes of data and require powerful GPUs to run efficiently. Their group models the initial molecular dynamic environment and predicted peptide structure first, and then perform the energy minimization to yield the initial conditions. MD simulations at this scale require substantial GPU resources, producing only 50-100 nanoseconds of simulation time per day on a single GPU.
Each simulation can take days and produce tens of gigabytes of output, making them impractical on standard computers. Access to HPC resources reduced project timelines from years to months, allowing the team to iterate quickly and move from computational design to experimental validation.
Their work reflects a broader vision for HPC-enabled science. Drawing on decades of experience with high-performance systems. Beyond accelerating timelines, HPC enables the team to explore more realistic biological scenarios, improving confidence in results before moving to costly lab experiments. The lab’s long-term goals include understanding what are the bioactive compounds found in food and how to create a novel bioactive compound for clinical applications, areas where computational scale directly translates into scientific impact.
While this project relied primarily on existing documentation for cluster access and operation, the team emphasizes thoughtful preparation as key to success. For researchers new to HPC, they recommend reading best practices, communicating with experts and staff, and planning ahead of your experiments. It is important to run a small pilot first before running a large job. Expect that there will be delays. Expect troubleshooting issues that are not easy. There are many tools for assisting and profiling runs. For instance, they mentioned that there are useful interactive tools such as Jupyter notebooks on the cluster for monitoring simulations and progress.
As Ilias notes, “We were able to complete simulations in about three months, work that would have taken years without HPC resources.”
Looking ahead, the team has recently completed experimental validation of their computational predictions in both cell lines and mouse models, with a publication expected in Spring 2026. Their work demonstrates how HPC@UCD enables researchers to move from AI-generated hypotheses to validated biological insight.
They used HPC resources, including the Farm and Hive clusters to perform molecular dynamics (MD) simulations (Gromacs software package) to understand how AMPs penetrate bacterial cell membranes to capture atomic-level detail that would be impossible to observe experimentally alone.