Creativity, Discovery, Engagement
Learning Like No Other
Materials Science and Biophysics Research Seminar Series
Stay up-to-date on the latest research and build relationships with academics at the Materials Science & Biophysics Seminar Series, which bring experts from around the world to campus to discuss their recent findings. Everyone is welcome!
Spring 2026 Seminar Series
|
Confusing Neural Networks in Simulational PhysicsFriday, January 23, 2026 Health Sciences Building EC 1218 2:30 - 3:30 PM Host: Dr. Trinanjan Datta (tdatta@augusta.edu) With the increasing pace of AI developments in corporate labs, questions about the future of academic groups in this field become louder. The most striking example in recent years might be AlphaFold, a machine-learning model that predicts the 3D protein structure from its amino-acid sequence and largely outperforms decade-long research in academia. This has led people to ask questions like “Are we (computational physicists) out of a job now?” Although understandable, I think this sentiment is too pessimistic. In this talk I willpresent two of our recent, student-led research projects where we use machine-learning tools and methods to analyze and inform conventional, statistical computer simulations. One of them is based on us purposely confusing neural networks, the other uses a quite confusion. |
|
|
Title to be announcedFriday, March 27, 2026 Health Sciences Building EC 1218 2:00 - 3:00 PM Host: Dr. Trinanjan Datta (tdatta@augusta.edu) Abstract to be announced ...
|
|
|
Computational Approaches for Proteins Beyond Experimental ResolutionFriday, April 17, 2026 Health Sciences Building EC 1218 2:00 - 3:00 PM Host: Dr. Trinanjan Datta (tdatta@augusta.edu) Deep learning-based protein structure prediction can reproduce experimentally determined tertiary structures with near-experimental accuracy for many globular proteins. However, model-reported confidence scores highlight persistent challenges for intrinsically disordered regions, conformational heterogeneity, and assemblies that are only partially structurally resolved. Moreover, structure prediction typically yields static structures and does not directly provide conformational populations or interaction energetics. In this seminar, I will present an integrative workflow used in our research group that couples deep learning-based structure prediction with all-atom molecular dynamics (MD) simulations to study protein systems beyond current experimental resolution. MD is used to explore conformational variability, test the stability of predicted structural features, and characterize protein-protein interfaces across an ensemble of states. The approach is illustrated through ongoing projects, including full-length modeling of the human prion protein (PrP) and protein complexes lacking experimentally determined structures. Finally, I will discuss post-simulation end-point analyses to estimate relative interaction energetics and identify residue-level contributions.
|
|
 Barrett Wells, PhD, Associate Dean for Life and Physical Sciences, Professor of Physics
|
Title to be announcedFriday, April 24, 2026 Health Sciences Building EC 1218 2:00 - 3:00 PM Host: Dr. Amani Jayakody (ajayakody@augusta.edu) Abstract to be announced ... |
|
Seminar Series Organizers
- Dr. Trinanjan Datta - tdatta@augusta.edu
- Dr. Shaobin Miao - smiao@augusta.edu
- Dr. Mustafa Culha - mculha@augusta.edu
- Dr. Amani Jayakody - ajayakody@augusta.edu
Seminar series sponsored by: Augusta University Research Institute, College of Science and Mathematics, Department of Physics and Biophysics