Inverse design of 3d molecular structures with conditional generative neural networks

  title={Inverse design of 3d molecular structures with conditional generative neural networks},
  author={Niklas W. A. Gebauer and Michael Gastegger and Stefaan S. P. Hessmann and Klaus-Robert M{\"u}ller and Kristof T. Sch{\"u}tt},
  journal={Nature Communications},
The rational design of molecules with desired properties is a long-standing challenge in chemistry. Generative neural networks have emerged as a powerful approach to sample novel molecules from a learned distribution. Here, we propose a conditional generative neural network for 3d molecular structures with specified chemical and structural properties. This approach is agnostic to chemical bonding and enables targeted sampling of novel molecules from conditional distributions, even in domains… 

Equivariant Shape-Conditioned Generation of 3D Molecules for Ligand-Based Drug Design

A new multimodal 3D generative model is introduced that enables shape-conditioned 3D molecular design by equivariantly encoding molecular shape and variationally encoding chemical identity, and ensures local geometric and chemical validity of generated molecules by using autoregressive fragment-based generation with heuristic bonding geometries.

RELATION: A Deep Generative Model for Structure-Based De Novo Drug Design.

The calculation results demonstrated that the RELATION model could efficiently generate novel molecules with favorable binding affinity and pharmacophore features and was used to design inhibitors for two targets, AKT1 and CDK2.

Structure-based drug design with geometric deep learning

An overview of the recent applications of geometric deep learning in bioorganic and medicinal chemistry highlights its potential for structure-based drug discovery and design, with emphasis on molecular property prediction, ligand binding site and pose prediction, and structure- based de novo molecular design.

Atomic structure generation from reconstructing structural fingerprints

This work implements this end-to-end structure generation approach using atom-centered symmetry functions as the representation and conditional variational autoencoders as the generative model and is able to successfully generate novel and valid atomic structures of sub-nanometer Pt nanoparticles as a proof of concept.

GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation

This paper proposes a novel generative model named GEODIFF for molecular conformation prediction that treats each atom as a particle and learns to directly reverse the diffusion process (i.e., transforming from a noise distribution to stable conformations) as a Markov chain.

Automatic Identification of Chemical Moieties

This work introduces a method to automatically identify chemical moieties (molecular building blocks) from atomic representations, enabling a variety of applications beyond property prediction, which otherwise rely on expert knowledge.

Cluster Generation via Deep Energy-Based Model

A new approach for the generation of stable structures of nanoclusters using deep learning methods, which can extrapolates the potential surface to cases of structures with larger number of atoms than was used in training, and finds the stable structures for the first time.

MolGenSurvey: A Systematic Survey in Machine Learning Models for Molecule Design

This paper systematically review the most relevant work in machine learning models for molecule design and summarizes all the existing molecule design problems into several venues according to the problem setup, including input, output types and goals.

Graph Neural Networks for Molecules

This review introduces GNNs and their various applications for small organic molecules and summarizes the recent development of self-supervised learning for molecules withGNNs.

3D pride without 2D prejudice: Bias-controlled multi-level generative models for structure-based ligand design

This work shows how, by deconvolving the generative posterior into chemical, topological and structural context factors, it not only avoid common pitfalls in the design and evaluation of generative models, but furthermore gain detailed insight into theGenerative process itself.



Generating equilibrium molecules with deep neural networks

A novel, autoregressive, convolutional deep neural network architecture that generates molecular equilibrium structures by sequentially placing atoms in three-dimensional space is introduced.

3DMolNet: A Generative Network for Molecular Structures

This work proposes a new approach to efficiently generate molecular structures that are not restricted to a fixed size or composition, based on the variational autoencoder which learns a translation-, rotation-, and permutation-invariant low-dimensional representation of molecules.

Inverse molecular design using machine learning: Generative models for matter engineering

Methods for achieving inverse design, which aims to discover tailored materials from the starting point of a particular desired functionality, are reviewed.

Symmetry-adapted generation of 3d point sets for the targeted discovery of molecules

A generative neural network for 3d point sets that respects the rotational invariance of the targeted structures is introduced and demonstrated its ability to approximate the distribution of equilibrium structures using spatial metrics as well as established measures from chemoinformatics.

Symmetry-Aware Actor-Critic for 3D Molecular Design

A novel actor-critic architecture for 3D molecular design that can generate molecular structures unattainable with previous approaches is proposed by exploiting the symmetries of the design process through a rotationally covariant state-action representation based on a spherical harmonics series expansion.


A novel probabilistic framework to generate valid and diverse conformations given a molecular graph, enjoying a high model capacity to estimate the multimodal conformation distribution and explicitly capturing the complex long-range dependencies between atoms in the observation space is proposed.

3D-Scaffold: Deep Learning Framework to Generate 3D Coordinates of Drug-like Molecules with Desired Scaffolds

The proposed 3D-Scaffold framework generates predominantly valid, unique, novel, and experimentally synthesizable molecules that have drug-like properties similar to the molecules in the training set and generalizes to new scaffolds, making it applicable to other domains.

3D-Scaffold: A Deep Learning Framework to Generate 3D Coordinates of Drug-like Molecules with Desired Scaffolds.

The deep learning model performs well with relatively small 3D structural training data and quickly learns to generalize to new scaffolds, highlighting its potential application to other domains for generating target specific candidates.

Molecular Geometry Prediction using a Deep Generative Graph Neural Network

A conditional deep generative graph neural network that learns an energy function by directly learning to generate molecular conformations that are energetically favorable and more likely to be observed experimentally in data-driven manner is proposed.

Accurate and transferable multitask prediction of chemical properties with an atoms-in-molecules neural network

With AIMNet, a modular and chemically inspired deep neural network potential, a new dimension of transferability is shown: the ability to learn new targets using multimodal information from previous training.