DeLinker - Deep Generative Models for 3D Linker Design

This repository contains our implementation of Deep Generative Models for 3D Linker Design (DeLinker).

If you found DeLinker useful, please cite our paper:

Imrie F, Bradley AR, van der Schaar M, Deane CM. Deep Generative Models for 3D Linker Design. Journal of Chemical Information and Modeling. 2020

@Article{Imrie2020,
author={Imrie, Fergus and Bradley, Anthony R. and van der Schaar, Mihaela and Deane, Charlotte M.},
title={Deep Generative Models for 3D Linker Design},
journal={Journal of Chemical Information and Modeling},
year={2020},
month={Mar},
day={20},
publisher={American Chemical Society},
issn={1549-9596},
doi={10.1021/acs.jcim.9b01120},
url={https://doi.org/10.1021/acs.jcim.9b01120}
}

Acknowledgements

We thank the authors of Constrained Graph Variational Autoencoders for Molecule Design for releasing their code. The code in this repository is based on their source code release (link). If you find this code useful, please consider citing their work.

Requirements

This code was tested in Python 3.6 with Tensorflow 1.10.

A yaml file containing all requirements is provided. This can be readily setup using conda.

conda env create -f DeLinker-env.yml
conda activate DeLinker-env

Data Extraction

Two primary datasets (ZINC and CASF) are in use.

To preprocess these datasets, please go to data directory and run prepare_data.py.

python prepare_data.py

Running DeLinker

We provide two settings of DeLinker. The first setting generates linkers with the same number of atoms as the reference molecule. The second setting generates linkers with a specified number of atoms.

To train and generate molecules using the first setting, use:

python DeLinker.py --dataset zinc --config '{"num_epochs": 10, "epoch_to_generate": 10, "train_file": "data/molecules_zinc_train.json", "valid_file": "data/molecules_zinc_valid.json"}'

To generate molecules with a pretrained model using the first setting, use

python DeLinker.py --dataset zinc --restore models/pretrained_DeLinker_model.pickle --config '{"generation": true, "number_of_generation_per_valid": 250, "batch_size": 1, "train_file": "data/molecules_zinc_test.json", "valid_file": "data/molecules_zinc_test.json"}'

To generate molecules using the second setting, use

python DeLinker_test.py --dataset zinc --restore models/pretrained_DeLinker_model.pickle --config '{"generation": true, "number_of_generation_per_valid": 250, "batch_size": 1, "train_file": "data/molecules_zinc_test_mode2.json", "valid_file": "data/molecules_zinc_test_mode2.json", "min_atoms": 3, "max_atoms": 11}'

In both cases, the output is of the following format:

Input fragments (SMILES) Ground truth molecule/fragments (SMILES) Generated molecule (SMILES)

More configurations can be found at function default_params in DeLinker.py.

Evaluation

A script to evaluate the generated molecules is provided in analysis directory.

python evaluate_generated_mols.py ZINC|CASF PATH_TO_GENERATED_MOLS PATH_TO_REFERENCE_MOLS ../data/data_zinc_final_train.txt SAVE_PATH OUTPUT_NAME NUM_CORES True None ./wehi_pains.csv >> log.txt

Pretrained Models and Generated Molecules

We provide a pretrained model:

models/pretrained_DeLinker_model.pickle

Generated molecules can be obtained upon request.

Examples

An example Jupyter notbook demonstrating the use of DeLinker for fragment linking can be found in the examples directory.

Contact (Questions/Bugs/Requests)

Please submit a Github issue or contact Fergus Imrie imrie@stats.ox.ac.uk.