RC_CVAE

Tensorflow implementation of the model described in the paper Conditional Variational AutoEncoder to Predict Suitable Conditions for Hydrogenation Reactions

Components

• descriptors_generation.py - functions for data preparation (x, y generation)
• cvae_models.py - functions for RC CVAE modeling
• Run_VAEs.ipynb - tutorial
• data/ - Dataset directory containing some data used in this study
• data/Reaxys_IDs/ - Reaxys® database identifiers for the test sets used in models evaluation.
• data/example_data/example_hydrogenation_USPTO.rdf - example dataset
• data/example_data/acids_bases_poisons_list.txt - list of additives: acids, bases, catalytic poisons
• data/example_data/catalysts_list.txt - list of catalysts

Data

Note: Due to the commercial nature of the Reaxys® database, we are unable to publicly release the complete datasets used in this study.

Reaxys® is a trademark of Elsevier Limited. Copyright © 2022 Elsevier Limited except certain content provided by third parties.

Instead, we provide Reaxys® IDs for the test sets (data/Reaxys_IDs/) and an example dataset of hydrogenation reactions from the USPTO database ("data/example_data/example_hydrogenation_USPTO.rdf") to enable code testing and validation. These hydrogenation reactions have been preliminarily standardized.

Usage

Example of usage can be found in Run_VAEs.ipynb

Dependencies

Only python 3.6

• numpy ~= 1.18.1
• tensorflow-gpu ~= 2.1.2 ; python_version == '3.6'
• tensorflow ~= 2.1.2
• tensorflow_probability ~= 0.9.0
• keras ~= 2.2.4
• h5py ~= 2.10.0
• git+https://github.com/cimm-kzn/s-vae-tf.git
• cgrtools == 3.1.9
• CIMtools == 3.1.0
• Fragmentor

Installation

1. Install dependencies from requirements.py
2. Download Fragmentor 2017 from https://github.com/cimm-kzn/CIMtools/tree/master/Fragmentor based on your OS and add it to bin/ folder of your virtual environment
3. Rename Fragmentor file to fragmentor-2017.x

Citation

@Article{Mazitov2025, title={Conditional variational autoEncoder to predict suitable conditions for hydrogenation reactions}, author={Mazitov, Daniyar A. and Poyezzhayeva, Assima and Afonina, Valentina A. and Gimadiev, Timur R. and Madzhidov, Timur I.}, }exa