https://github.com/Frank-LIU-520/DeepMoleNet

Deep learning for molecules quantum chemistry properties prediction
https://github.com/Frank-LIU-520/DeepMoleNet

deep-learning graph-convolutional-neural-networks quantum-chemistry quantum-machine-learning

Last synced: 3 months ago
JSON representation

Deep learning for molecules quantum chemistry properties prediction

• Host: GitHub
• URL: https://github.com/Frank-LIU-520/DeepMoleNet
• Owner: Frank-LIU-520
• License: mpl-2.0
• Created: 2021-01-03T09:01:41.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-04-14T04:33:44.000Z (over 3 years ago)
• Last Synced: 2024-06-29T12:34:59.518Z (5 months ago)
• Topics: deep-learning, graph-convolutional-neural-networks, quantum-chemistry, quantum-machine-learning
• Language: Python
• Homepage:
• Size: 43 KB
• Stars: 35
• Watchers: 2
• Forks: 6
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# DeepMoleNet

  DeepMoleNet is a deep learning package for molecular properties prediction. It is developed with a multi-level attention mechanism to enable chemical interpretable insights being fused into multi-task learning through (1) weighting contributions from various atoms and (2) taking the atom-centered symmetry functions (ACSFs) as the teacher descriptor, rather than using ACSFs as input in the conventional way. The properties including dipole moment, HOMO, and Gibbs free energy within chemical accuracy are achieved by using multiple benchmarks, both at the equilibrium and non-equilibrium geometries.

  Our model is also the champion solution of Codalab Alchemy Contest with the former name ape-MPNN. https://alchemy.tencent.com/

      To cite this algorithm, please reference: Ziteng Liu, Liqiang Lin, Qingqing Jia, Zheng Cheng, Yanyan Jiang, Yanwen Guo*, Jing Ma*. "Transferable multilevel attention neural network for accurate prediction of quantum chemistry properties via multitask learning." JCIM 2021 https://doi.org/10.1021/acs.jcim.0c01224   https://pubs.acs.org/doi/10.1021/acs.jcim.0c01224#

      

After download the code, one can use DeepMoleNet model in the following steps:

    1. To unzip the download zip file and make sure that all packages (including ase, networkx, torch, torch-cluster, torch-geometric, torch-scatter, rdkit, pathlib, pandas, numpy, dscribe) listed in the requirement.txt file are installed.

    

    2. To put molecule sdf files(QM9, Alchemy, MD17, ANI-1ccx and etc) in the .\data-bin\raw\dev, with its sdf file name and all 12 properties saved in .\data-bin\raw\dev_target.csv in templet.

    

    3. To run the code, python DeepMoleNet.py

We also provide trained models of single-target and multi-target results trained on QM9, models trained on MD17 and models trained on ANI-1ccx for reporoducibility and other uses like transfer learning on small dataset.

Please feel free to contact us via [email protected] for any questions and visit the following website for details and updates in our codes.

https://itcc.nju.edu.cn/majing/software_en.html