Generalized Biomolecular Modeling and Design with RoseTTAFold All-Atom

Explore the advancements in generalized biomolecular modeling and design through this comprehensive talk on RoseTTAFold All-Atom (RFAA) and RFdiffusion All-Atom (RFdiffusionAA). Delve into the capabilities of these deep learning networks in modeling complex biological assemblies, including proteins, nucleic acids, small molecules, metals, and covalent modifications. Learn about the architectural changes made to RoseTTAFold 2, dataset augmentation techniques, and the loss function used in training. Examine the structure prediction results and compare RFAA's performance to AlphaFold2 in various scenarios. Discover the process of backbone generation and in silico results, followed by experimental validation of designs for binding pockets of therapeutic molecules and enzymatic cofactors. Gain insights into the potential applications of RFAA and RFdiffusionAA in modeling and designing complex biomolecular systems, with a focus on expanding the range of wavelengths captured by photosynthesis.

- Intro
- RoseTTAFold All-Atom
- Architectural changes to RoseTTAFold 2
- Dataset augmentation
- Loss function
- Structure prediction results
- Backbone generation
- In silico results
- Experimental validation of designs
- Q&A