search

chaidiscovery / chai-lab

Chai-1, SOTA model for biomolecular structure prediction
https://www.chaidiscovery.com
Other
1k stars 130 forks source link

Whether the ligand supports metal ions #28

Open Gibo-Chun opened 1 week ago

Gibo-Chun commented 1 week ago

Thank you for developing chai-lab, which is an amazing tool and has greatly assisted my work. However, when I try to input the trivalent iron ion in the ligand part locally using chai_lab with the command >ligand|FE3P [Fe+3] I encounter an error: [22:51:19] UFFTYPER: Unrecognized atom type: Fe2+2 (0) Does this mean that chai-lab does not support trivalent iron ions?

arogozhnikov commented 1 week ago

Hi Lancelot, We rely on rdkit, and am a bit surprised with this warning.

But maybe rdkit's approach is the best here - I'd not expect trivalent ions to exist in 'normal conditions'

duerrsimon commented 1 week ago

@arogozhnikov for protein structure prediction Fe3+ is quite important. Many very important proteins like ferredoxin or cytochromes contain Fe3+ (either bound to a cofactor or directly by the protein). The ability of iron to switch oxidation state from Fe2+ to Fe3+ is absolutely essential for life. In fact the default oxidation state for iron in the protein databank (3-letter code FE) is also Fe3+ https://www.rcsb.org/ligand/FE

Would be great to support Fe3+ too. For me Fe3+ predictions work

duerrsimon commented 1 week ago

@Gibo-Chun I just tried running to predict Fe3+ and for me it works:

>ligand|and-example-for-ligand-encoded-as-smiles
[Fe+3]
arogozhnikov commented 1 week ago

Hi Simon, thanks for explanation and for checking!

rdkit repo has related discussion: https://github.com/rdkit/rdkit/discussions/7119