Chirality of Ligand Not Preserved

[rjrich]

When I predict the structures of protein-ligand complexes using Boltz, the chirality of the input ligands is not being preserved. Some of the ligands have a chiral phosphorus atom and others have a chiral carbon atom. In addition, organophosphorus compounds that have a P=S structure are being changed to a P-S-H structure in the output complexes. I am inputting the ligand files in v3000 SDF format, and the SDF files were prepared using Schrodinger Maestro, which uses RDKit to generate the SDF files. Is there any way to preserve the chirality and/or the P=S moiety of the input ligands when predicting protein-ligand complexes using Boltz?

I believe that this problem stems from the fundamental issue of attempting to convey 3D information of ligands using 1D SMILES strings rather than a format that contains 3D coordinates. Why is it that AF3 reproductions currently must use SMILES strings or CCD codes as input for ligands?

Thank you.