Internal-Coordinate Density Modelling of Protein Structure: Covariance Matters
Not on the map. Why?
Abstract (arXiv)
After the recent ground-breaking advances in protein structure prediction, one of the remaining challenges in protein machine learning is to reliably predict distributions of structural states. Parametric models of fluctuations are difficult to fit due to complex covariance structures between degrees of freedom in the protein chain, often causing models to either violate local or global structural constraints. In this paper, we present a new strategy for modelling protein densities in internal coordinates, which uses constraints in 3D space to induce covariance structure between the internal degrees of freedom. We illustrate the potential of the procedure by constructing a variational autoencoder with full covariance output induced by the constraints implied by the conditional mean in 3D, and demonstrate that our approach makes it possible to scale density models of internal coordinates to full protein backbones in two settings: 1) a unimodal setting for proteins exhibiting small fluctuations and limited amounts of available data, and 2) a multimodal setting for larger conformational changes in a high data regime.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 12 of its 61 references that are in the map. First posted on arXiv 27 Feb 2023.
Builds on
12 of 61 references.
References · 12
- 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
- 2022Torsional Diffusion for Molecular Conformer Generationcited by 263
- 2017VAMPnets: Deep learning of molecular kineticscited by 136
- 2022Protein structure generation via folding diffusioncited by 115
- 2022Protein Structure and Sequence Generation with Equivariant Denoising Diffusion Probabilistic Modelscited by 98
- 2023Two for One: Diffusion Models and Force Fields for Coarse-Grained Molecular Dynamicscited by 72
- 2021Deeptime: a Python library for machine learning dynamical models from time series datacited by 38
- 2022Flow-matching -- efficient coarse-graining of molecular dynamics without forcescited by 36
- 2017Variational Encoding of Complex Dynamicscited by 21
- 2021Inferring a Continuous Distribution of Atom Coordinates from Cryo-EM Images using VAEscited by 17
- 2019Graph Dynamical Networks for Unsupervised Learning of Atomic Scale Dynamics in Materialscited by 13
- 2018Transferable neural networks for enhanced sampling of protein dynamicscited by 9
Built on by
Citing papers per year: 2025: 1.
1 paper.
- 2025AlphaFold's Bayesian Roots in Probability Kinematicscited by 1
- 2025AlphaFold's Bayesian Roots in Probability Kinematics26 May 2025