How to generate 10000 molecules on a protein?

[liutx08]

I'm currently reading your paper and learning how to run the code. Here I have encountered a small question: How to generate many molecules on a protein target? For example if I want to generate 10000 molecules on protein 1e9u using the script beamsize.py, is it the beamsize argument that I should change? Or there are other ways of doing so? Looking forward to your reply!

[BL-Lac149597870]

Hello @liutx08 , for a straightforward approach, consider executing the beamsize.py script 1000 times with the beam size set at 10. If preferred, you have the option to increase the beam size to 20, in which case the script would need only 500 runs. However, it's important to keep in mind that a larger beam size might lead to a higher likelihood of generating invalid SMILES. The choice of configuration depends on your specific requirements and constraints.

[liutx08]

Thanks for the reply! I tried to run beamsearch.py for 10 times setting beam size at 10, however I noticed that the SMILES generated each time are the same. Is there arguments or switches for random seed settings?