WGAN - Molecular - Graphs Open - Source Model: Accelerating Drug Discovery and Generating Novel Molecular Graph Structures

Wgan Molecular Graphs

Developed by keras-io

This model uses WGAN-GP and R-GCN architectures to generate novel molecular graph structures, accelerating the drug discovery process.

Molecular Model #Molecular Generation #WGAN-GP Optimization #R-GCN Architecture

Downloads 28

Release Time : 6/11/2022

Model Overview

This is a generative adversarial network model specifically designed for generating novel molecular graph structures. By learning the properties of known molecules, the model can explore vast chemical spaces and generate potentially valuable drug candidate molecules.

Model Features

Accelerated Drug Discovery

Significantly reduces the time and cost of traditional drug discovery processes by generating novel molecular structures.

WGAN-GP Architecture

Uses an improved Wasserstein GAN architecture with gradient penalty to enhance training stability.

R-GCN Support

Employs Relational Graph Convolutional Networks to process molecular graph structure data.

QM9 Dataset Optimization

Specifically optimized for the QM9 quantum mechanics dataset, making it suitable for entry-level molecular generation.

Model Capabilities

Molecular Graph Generation

Novel Molecular Structure Exploration

Drug Candidate Molecule Prediction

Use Cases

Drug Development

Drug Candidate Molecule Generation

Generates novel molecular structures with potential medicinal value.

Explores molecular spaces difficult to discover using traditional methods.

Molecular Property Prediction

Predicts properties of generated molecules such as solubility, toxicity, and protein affinity.

Chemical Research

Novel Compound Exploration

Automatically explores possible compound structures within vast chemical spaces.

🚀 tf-keras

This repository contains a generative model for graphs, which can be used to generate novel molecules, addressing the time - consuming and costly process of new drug development.

🚀 Quick Start

This repo provides a generative model for graphs and a notebook to implement it. You can refer to the code and documentation here to generate novel molecules.

✨ Features

Implement a generative model for graphs to generate novel molecules.
Use the WGAN - GP with R - GCN architecture for small molecular graph generation.

📚 Documentation

Model description

This repo contains the model and the notebook for implementing a generative model for graphs and using it to generate novel molecules WGAN - GP with R - GCN for the generation of small molecular graphs.

Full credits go to Alexander Kensert

Reproduced by Vu Minh Chien

Motivation: The development of new drugs (molecules) can be extremely time - consuming and costly. The use of deep learning models can alleviate the search for good candidate drugs, by predicting the properties of known molecules (e.g., solubility, toxicity, affinity to the target protein, etc.). As the number of possible molecules is astronomical, the space in which we search for/explore molecules is just a fraction of the entire space. Therefore, it's arguably desirable to implement generative models that can learn to generate novel molecules (which would otherwise have never been explored).

Intended uses & limitations

Recent implementations of generative models for molecular graphs also include Mol - CycleGAN, GraphVAE, and JT - VAE. For more information on generative adversarial networks, see GAN, WGAN, and WGAN - GP.

Training and evaluation data

The dataset used in this tutorial is a quantum mechanics dataset (QM9), obtained from MoleculeNet. The QM9 dataset is a good first dataset to work with for generating graphs, as the maximum number of heavy (non - hydrogen) atoms found in a molecule is only nine.

Training hyperparameters

The following hyperparameters were used during training:

generator_learning_rate: 5e - 04
discriminator_learning_rate: 5e - 04
train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
num_epochs: 50

Training results

The results are shown in TensorBoard (Training metrics).

Model Plot

View Model Demo

Model Demo

View Model Plot

Model Image

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご