Deep Generative Design of Porous Organic Cages via a Variational Autoencoder

This repository contains data and codes for Cage-VAE in our paper Deep generative design of porous organic cages via a variational autoencoder.

Description

In this work, we present a VAE model, Cage-VAE, for the targeted generation of shape-persistent porous organic cages (POCs). We demonstrate the capacity of Cage-VAE to propose novel, shape-persistent POCs, via integration with multiple efficient sampling methods, including bayesian optimization and spherical linear interpolation.

Contents

• Data

  • Please unzip datasets.zip and put original_original.csv and dataset_augmented.csv to ./datasets/ directory.

• Main

  • dataset_analysis.py: Overview of the original dataset.
  • data_augmentation.py: Methods used to create the augmented dataset.
  • training.py: Training of a new Cage-VAE.
  • model_eval.py: Integrated evaluations of an existing Cage-VAE.
  • conditional_generation.py: conditional generation of shape-persistent POCs using sampling methods.

• Utils

  • analysis_utils: import as a_utils, useful tools for the analysis of the generative model.
  • encoding_utils: import as e_utils, useful tools for molecular encodings and moiety modifications.
  • VAE: import as VAE, VAE architecture and methods.
  • generation_utils:import as g_utils, useful tools for molecular generations.
  • utils: import as utils, other utils.

• Results

  • ./cage/: folder contains samples of generated POCs

Requirements

$ conda create --name cage-VAE python=3.7.11
$ conda activate cage-VAE
# install requirements
$ pip install numpy==1.21.5
$ pip install pandas==1.3.3
$ pip install matplotlib==3.2.1
$ pip install seaborn==0.11.2
$ pip install scikit-learn==0.22.1
$ pip install torch==1.9.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html
$ pip install torchvision==0.2.2
$ conda install -c conda-forge rdkit=2020.09.1.0 
$ pip install stk==2021.8.2.0
$ pip install selfies==2.0.0
$ pip install GPyOpt==1.2.6
$ conda install -c conda-forge/label/cf202003 openbabel=3.0.0
$ conda install -c conda-forge pymatgen=2022.1.7

For running in an interactive window

$ pip install ipykernel==6.4.1