Generating valid Euclidean distance matrices

By Moritz Hoffmann, Frank No'e. 2019

58 citations recorded by the source corpus.

Generating point clouds, e.g., molecular structures, in arbitrary rotations, translations, and enumerations remains a challenging task. Meanwhile, neural networks utilizing symmetry invariant layers have been shown to be able to optimize their training objective in a data-efficient way. In this spirit, we present an architecture which allows to produce valid Euclidean distance matrices, which by construction are already invariant under rotation and translation of the described object. Motivated by the goal to generate molecular structures in Cartesian space, we use this architecture to construct a Wasserstein GAN utilizing a permutation invariant critic network. This makes it possible to generate molecular structures in a one-shot fashion by producing Euclidean distance matrices which have a three-dimensional embedding.

Subfields: Computer Vision and Pattern Recognition, Materials Chemistry, Ocean Engineering

How researchers see this paper

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

In Computer Vision and Pattern Recognition

1. Jian Peng
2. Jian Tang
3. Minkai Xu
4. Shitong Luo
5. Yoshua Bengio
6. Shuangjia Zheng
7. Luc Van Gool
8. Arash Vahdat

In Materials Chemistry

1. Jian Tang
2. Minkai Xu
3. Shitong Luo
4. Jian Peng
5. Jiahua Rao
6. Yoshua Bengio
7. Shuangjia Zheng
8. Tommi S. Jaakkola

In Ocean Engineering

1. Yoshua Bengio

Citations in the corpus

• TANKBind: Trigonometry-Aware Neural NetworKs for Drug-Protein Binding Structure Prediction (2022, background): “The Cartesian coordinates, {x̂j}, of the heavy atoms of a drug compound could be deduced analytically based on the predicted inter-molecular distance matrix, D ij , the coordinates of protein nodes, {xpi }, and the pair distance matrix of compound nodes, D jk′ [46, 47].”
• Learning Neural Generative Dynamics for Molecular Conformation Generation (2021, background): “Some other works (Lemke & Peter, 2019; AlQuraishi, 2019; Ingraham et al., 2019; No´e et al., 2019; Senior et al., 2020) also focus on learning models to directly generate 3D structure, but focus on the protein folding problem.”