search

lilleswing / jt-vae

Junction Tree Variational AutoEncoder Implementation Attempt
MIT License
10 stars 1 forks source link

A little question about tree decomposition algorithm. #1

Open limaosen0 opened 6 years ago

limaosen0 commented 6 years ago

Dear Dr. Leswing: I learn and run your code "tree_decomp.py", and I found a little mistake, where it is at line 219. We should generate a new SMILES but this is still the operation of mol and return a mol. Thus I try to change it to be "substructure_smiles = [Chem.MolToSmiles(x) for x in substructures]". But the generated two .csv are the same, as well as the two .json files. We cannot get the results like your uploaded files in '/data'. I am a new learner of rdkit and may ask some fool questions, please forgive me and help me. Thank you very much.

lilleswing commented 6 years ago

Hi limasen0 this was my attempt to reproduce the paper, since then the original code was published here. These attempts were unfinished since the original source was published. https://github.com/wengong-jin/icml18-jtnn

I have made small changes (most noteably python3 compatibility) here. https://github.com/lilleswing/icml18-jtnn

See the actual MolTree code from the paper https://github.com/lilleswing/icml18-jtnn/blob/master/jtnn/mol_tree.py

limaosen0 commented 6 years ago

Thank you very much!