Differences between sample.py and sample_for_pdb.py

[VincentBt]

Hi, thank you for the great piece of work!

I was comparing sample.py and sample_for_pdb.py and I noticed that only the second file had queue_weight. Is there any reason for that? I guess this helps generating more diverse molecules.

I also noticed another difference: the possibility in sample.py (but not in sample_for_pdb.py) to use LigandBFSMask masking instead of LigandMaskAll: what is the use of it?

Thank you very much!

[pengxingang]

Hi, thanks for your interest.

The queue_weight parameter was introduced to address situations where a single sample has an excessive number of descendant samples. This can lead to decreased diversity in the generated molecules. Moreover, we have noticed that it can sometimes result in the generation of molecules with multiple rings, resembling polycyclic aromatic hydrocarbons, which is not desired.

Regarding the LigandBFSMask and LigandMaskAll options for sampling, if the masking ratio is set to 100%, there is no practical difference between the two. You can choose either option without affecting the outcome.