Pushing the frontiers of classical structural biology with AI

Our research group, comprised of dedicated CS and BioScience students and esteemed faculty, is at the forefront of exploring groundbreaking advancements in ML/AI to push the frontiers of what we can achieve with classical structural biology data. In particular, this group focuses on problems like the determination of protein structures at an atomic level, which remains a significant challenge in structural biology, despite decades of advancements. X-ray crystallography is a powerful tool in many aspects of science and engineering, particularly in materials science, chemistry and the biological sciences, where supplemental experiments are needed to solve new structures. Yet, such traditional approaches encounter notable obstacles such as the unresolved crystallographic phase problem. Our goal is to introduce a family of models, also known as the CrysFormer family of models, that are novel hybrid models that exploit the strengths of (visual) transformers with the aim of integrating experimental and ML approaches to protein structure determination from crystallographic data.

Through rigorous research and collaboration, we strive to contribute to the rapidly evolving field of Structural Biology + AI. Our projects are designed not only to advance practical understanding but also to develop algorithmic solutions that can be applied across various domains. We invite you to explore our website to learn more about our initiatives, ongoing projects, and the impact of our research. Join us on this exciting journey as we unlock new dimensions of computing and pave the way for future innovations in technology and science.

The results of this project include:

• XYZ.
• XYZ.
• XYZ.

Publications

Tom Pan, Shikai Jin, Mitchell D Miller, Anastasios Kyrillidis, George N Phillips, ``A deep learning solution for crystallographic structure determination’‘, International Union of Crystallography, IUCrJ, Volume 10, Issue 4, Pages 487-496, 2023.

Qiutai Pan, Chen Dun, Shikai Jin, Mitchell D. Miller, Anastasios Kyrillidis, George N. Phillips Jr. ``CrysFormer: Protein Structure Determination via Patterson Maps, Deep Learning and Partial Structure Attention’‘, Accepted at Structural Dynamics 2024.

Qiutai Pan, Evan Dramko, Mitchell D. Miller, George N Phillips Jr., Anastasios Kyrillidis, ``RecCrysFormer: Refined protein structural prediction from 3D Patterson maps via recycling training runs’‘, Submitted to top-ML conference, 2024.