Deep generative models of 3D grids for structure-based drug discovery

Last update: Jan 03, 2023

Overview

What is liGAN?

liGAN is a research codebase for training and evaluating deep generative models for de novo drug design based on 3D atomic density grids. It is based on libmolgrid and the gnina fork of caffe.

VAE paper - 2 minute talk

CVAE paper - 15 minute talk

Dependencies

numpy
pandas
scikit-image
openbabel
rdkit
molgrid
torch
protobuf
gnina version of caffe

Usage

You can use the scripts download_data.sh and download_weights.sh to download the test data and weights that were evaluated in the above papers.

The script generate.py is used to generate atomic density grids and molecular structures from a trained generative model.

Its basic usage can be seen in the scripts generate_vae.sh:

LIG_FILE=$1 # e.g. data/molport/0/102906000_8.sdf

python3 generate.py \
  --data_model_file models/data_48_0.5_molport.model \
  --gen_model_file models/vae.model \
  --gen_weights_file weights/gen_e_0.1_1_disc_x_10_0.molportFULL_rand_.0.0_gen_iter_100000.caffemodel \
  --rec_file data/molport/10gs_rec.pdb \
  --lig_file $LIG_FILE \
  --out_prefix VAE \
  --n_samples 10 \
  --fit_atoms \
  --dkoes_make_mol \
  --output_sdf \
  --output_dx \
  --gpu

And generate_cvae.sh:

REC_FILE=$1 # e.g. data/crossdock2020/PARP1_HUMAN_775_1012_0/2rd6_A_rec.pdb
LIG_FILE=$2 # e.g. data/crossdock2020/PARP1_HUMAN_775_1012_0/2rd6_A_rec_2rd6_78p_lig_tt_min.sdf

python3 generate.py \
  --data_model_file models/data_48_0.5_crossdock.model \
  --gen_model_file models/cvae.model \
  --gen_weights_file weights/lessskip_crossdocked_increased_1.lowrmsd.0_gen_iter_1500000.caffemodel \
  --rec_file $REC_FILE \
  --lig_file $LIG_FILE \
  --out_prefix CVAE \
  --n_samples 10 \
  --fit_atoms \
  --dkoes_make_mol \
  --output_sdf \
  --output_dx \
  --gpu

Both scripts can be run from the root directory of the repository.

Deep generative models of 3D grids for structure-based drug discovery

Related tags

Overview

What is liGAN?

Dependencies

Usage

Owner

Matt Ragoza

Official PyTorch implementation of "VITON-HD: High-Resolution Virtual Try-On via Misalignment-Aware Normalization" (CVPR 2021)

A simplified framework and utilities for PyTorch

Code for the paper SphereRPN: Learning Spheres for High-Quality Region Proposals on 3D Point Clouds Object Detection, ICIP 2021.

C3DPO - Canonical 3D Pose Networks for Non-rigid Structure From Motion.

KITTI-360 Annotation Tool is a framework that developed based on python(cherrypy + jinja2 + sqlite3) as the server end and javascript + WebGL as the front end.

Code for our ALiBi method for transformer language models.

Code release for our paper, "SimNet: Enabling Robust Unknown Object Manipulation from Pure Synthetic Data via Stereo"

Pytorch implementation of AngularGrad: A New Optimization Technique for Angular Convergence of Convolutional Neural Networks

Open-source implementation of Google Vizier for hyper parameters tuning

PyTorch implementation of "Debiased Visual Question Answering from Feature and Sample Perspectives" (NeurIPS 2021)

Reimplementation of Learning Mesh-based Simulation With Graph Networks

Repository for "Improving evidential deep learning via multi-task learning," published in AAAI2022

Elevation Mapping on GPU.

A series of convenience functions to make basic image processing operations such as translation, rotation, resizing, skeletonization, and displaying Matplotlib images easier with OpenCV and Python.

FairEdit: Preserving Fairness in Graph Neural Networks through Greedy Graph Editing

ICCV2021 - Mining Contextual Information Beyond Image for Semantic Segmentation

Playing around with FastAPI and streamlit to create a YoloV5 object detector

Source code release of the paper: Knowledge-Guided Deep Fractal Neural Networks for Human Pose Estimation.

Classification Modeling: Probability of Default

Lipschitz-constrained Unsupervised Skill Discovery