Home

Awesome

NMR-TS

Molecule identification from NMR spectrum using de novo molecule generator.

This code is for reference only as it requires MPI and Gaussian environment which is hard to ensure among different clusters. Also, for the NMR spectrum prediction part, we have used a Gaussain encapsulation package which we can not make public for now. Replacing this part by a traditional Gaussian code or other NMR spectrum prediction method would solve the problem.

Requirements

  1. Python 2.7.15

  2. RDKit '2018.09.1'

  3. TensorFlow 1.9.0

  4. Gaussian

How to use?

Clone repo.

git clone https://github.com/tsudalab/NMR-TS.git
cd NMR-TS

Train the RNN model.*

cd train_RNN
python train_RNN.py

*You need a SMILES dataset for this step

Run main program.

python mpi_thread_ChemTS_tree_vl_jz.py

Interface program of Gaussian

To run the main program, it is necessary to implement the interface that connect the program with a simulation program of NMR whose input is sdf file. If you are interested in the interface program between the program and Gaussian, please contact masato.sumita@riken.jp (Masato Sumita Ph. D., http://www17.plala.or.jp/nymphea/index.html).

License

This project is licensed under the terms of the MIT license.