Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Incorrect Aromaticity and Planarity Interpretation for Bicyclic Molecule #65

Open
Jnelen opened this issue Nov 28, 2024 · 2 comments
Open

Comments

@Jnelen
Copy link

Jnelen commented Nov 28, 2024

I have encountered a problem with the way the software interprets the structure of a bicyclic molecule, leading to incorrect results in the output. Specifically, the molecule in question consists of a bicyclic structure where only one of the rings is aromatic. However, during processing, both rings are incorrectly interpreted as aromatic. Additionally, the non-aromatic ring is also treated as planar, which is not chemically accurate and affects downstream results when working with this ligand in complex with a protein.

For example, consider the structure of the input molecule shown below:
image

When this ligand is processed with the protein, the output structure incorrectly identifies both rings as aromatic and planar. You can see my result here:
image

To reproduce the issue, you can use the attached Fasta file along with the ligand structure provided in either the regular aromatic SMILES format or the kekule format (explicit double bonds). I have tried both, but the results remain the same in both cases. The input and output files are included in the attached zip file for your convenience: boltz_aromaticity.zip

@gcorso
Copy link
Collaborator

gcorso commented Nov 29, 2024

Thank you for sharing this finding @Jnelen!

Currently, the model (similarly to AlphaFold3) does not take as input information about whether the bonds are single/double/aromatic so changing the input is not indeed supposed to change the output.

We are experimenting with different ways of fine-tuning the model to solve some of these issues and will share improvements as soon as we find them!

@Jnelen
Copy link
Author

Jnelen commented Dec 2, 2024

Thank you for sharing this finding @Jnelen!

Currently, the model (similarly to AlphaFold3) does not take as input information about whether the bonds are single/double/aromatic so changing the input is not indeed supposed to change the output.

We are experimenting with different ways of fine-tuning the model to solve some of these issues and will share improvements as soon as we find them!

Thanks a lot for the clarification! Looking forward to seeing the future improvements and evolution of this exciting project!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants