Random sampled 0 / 100 AND Docking failed

[xukm3]

Hi, I am a doctoral student in China with a biomedical background. I have found your outstanding work in the drug screening work. I have encountered some problems in the installation and operation of the program.

In the previous work, I carried out virtual screening, but the free binding energy of the screened inhibitors was larger than that of the natural protein ligands. Therefore, as you said in your article, the size of the compound library may determine whether enough good compounds can be screened. Therefore, I want to try to conduct de novo drug design on this basis. After installing the environment dependencies, the three attached jupyter notebook files are now running smoothly. Let's see if we can run it on the sample. I'm using wsl, using the __python generate_molecules.py gvae -o./outputdir, Some outputs then appear including [ERROR] 2023-07-25 16:05:15,085 Failed parsing for C1=C(Br)C=CC@HC(=CN2)/C=C3\C(=O)N(C)C(=N[H])\N3C_

I think it's the Br atom, but the problem doesn't affect the program. random sampled 0 and Docking failed failed to produce output. I would like to ask for your advice. I wish you good health and happiness.

[xukm3]

Hello？

[cieplinski-tobiasz]

Hi, Thanks for your interest in our paper.

You are right, although the parsing failed, model fine tuning continues. This is because if we cannot parse a SMILES during model fine tuning, we just discard it and fine tune it with valid SMILES.

Regarding the second part. It is a bit misleading, but generation proceeds as expected. In this scenario GVAE either generated invalid SMILES, or generated SMILES failed to dock, probably due to timeout. Unfortunately, we also encountered such problems with this model, and generation took us a lot of time.

Please try to generate molecules using the CVAE model. It used to produce more valid molecules. If it still fails, please reach out.

[xukm3]

Thanks for your kindly reply. According to your suggestion, to use CVAEmodel, the instructions used are: python generate_molecules.py cvae -o./output_dir -n 5 -f 1 After running, the following information is displayed, and Random sampled 0/100 continues to appear.