Clarification regarding the number of molecular building blocks. Why they are different from JT-VAE?

[Srilok]

Hello,

First, I really enjoyed reading the paper. Amazing work!

I have a question regarding the number of building blocks used for generating small molecules. Appendix A.3 of the paper states that there are a total of 105 unique building blocks (after accounting for different attachment points) and that they were obtained by the process suggested by the JT-VAE paper. (Jin et al. (2020)). However, in the JT-VAE paper, the total vocabulary size is $|\chi|=780$ obtained from the same ZINC dataset. My understanding is they are both the same. If that is correct, why are the number of building blocks different here? What am I missing? If they are not the same, can you please explain the difference?

Thank you so much for your help

[MKorablyov]

The building blocks in two papers are not the same but quite similar. In both cases we represent molecules as junction trees - that means there are no cycles. Ours are obtained by BRICS followed by Bemis-Murcko decomposition. Finally, we had a chemist who curated our set of building blocks. In the end, I think our building blocks ended up slightly smaller and more rigid compared to JT-VAE and worked better for us in practice.

BRICS: https://www.rdkit.org/docs/source/rdkit.Chem.BRICS.html
Bemis-Murcko: https://rdkit.org/docs/source/rdkit.Chem.Scaffolds.MurckoScaffold.html

[Srilok]

Thank you. After performing the BRICS followed by Bemis-Murcko decomposition on the 250k SMILES dataset, I get 8962 unique building blocks. Can you please comment a bit more about the curation process? How did you narrow down to a smaller list of 105 building blocks?

Also, how did you determine the attachment points (block_r in data/blocks_PDB_105.json)?

Thank you

[Srilok]

Hi @MKorablyov , just following up on my comment earlier. It would be really helpful if you could provide those details. Thank you so much!