Protein structure generation via folding diffusion

By Kevin E. Wu, Kevin Kaichuang Yang, Rianne van den Berg, James Y. Zou, Alex X. Lu, Ava P. Amini. 2022

297 citations recorded by the source corpus.

The ability to computationally generate novel yet physically foldable protein structures could lead to new biological discoveries and new treatments targeting yet incurable diseases. Despite recent advances in protein structure prediction, directly generating diverse, novel protein structures from neural networks remains difficult. In this work, we present a diffusion-based generative model that generates protein backbone structures via a procedure inspired by the natural folding process. We describe a protein backbone structure as a sequence of angles capturing the relative orientation of the constituent backbone atoms, and generate structures by denoising from a random, unfolded state towards a stable folded structure. Not only does this mirror how proteins natively twist into energetically favorable conformations, the inherent shift and rotational invariance of this representation crucially alleviates the need for more complex equivariant networks. We train a denoising diffusion probabilistic model with a simple transformer backbone and demonstrate that our resulting model unconditionally generates highly realistic protein structures with complexity and structural patterns akin to those of naturally-occurring proteins. As a useful resource, we release an open-source codebase and trained models for protein structure diffusion.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Arash Vahdat
2. Tero Karras
3. Yoshua Bengio
4. Richard Zhang
5. Yang Song
6. Jun-Yan Zhu
7. Luc Van Gool
8. Axel Sauer

Citations in the corpus

• Broadly applicable and accurate protein design by integrating structure prediction networks and diffusion generative models (2022, result): “This definition of success is significantly more stringent than those described elsewhere (refs [5,8,16,25], Fig.”
• A Survey on Generative Diffusion Models (2022, background): “Alternatively, generating models that capture multiple features and the distribution of structural features in high-dimensional space can lead to more diverse distributionsandinterpretability[199],[200],[201],[202].”
• eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers (2022, background): “They have driven successful applications such as text-to-image generation [57, 59, 63], natural language generation [41], time series prediction [76], audio synthesis [39], 3D shape generation [49, 87, 91], molecular conformation generation [84], protein structure generation [83], machine learning s”
• Illuminating protein space with a programmable generative model (2022, unknown): “24,25), or cubically, O ( N 3 ) (refs.”
• SE(3) diffusion model with application to protein backbone generation (2023, background): “Past works have developed diffusion models over different representations of protein structures without pretraining (Wu et al., 2022; Trippe et al., 2022; Anand & Achim, 2022; Qiao et al., 2022; Ingraham et al., 2022).”