The results ligand is different from SMILES I have entered

[MinsikBae]

I am experiencing a discrepancy between the ligand structure I expect based on my input SMILES and the 3D structure generated by Boltz. My SMILES string specifies an alkane chain (CCCCCCCN) as a linker, but the output structure has an alkene (C=C) in that linker(1 to 3 alkene generated), and the molecule forms an unexpected macrocycle.

I have already attempted the following troubleshooting steps:

Used the MMFF94 force field instead of UFF.
Modified the compute_3d_conformer function in schema.py to check for and convert unexpected double bonds to single bonds.
Verified that useRandomCoords is effectively disabled.
Checked input SMILES string
Adjust recycling_steps , sampling_steps, step_scale

[JesusSeco]

I had a very similar problem. Using canonical SMILES worked, but the most effective solution, is to turn off in schema.py the Sanitize option to False, to avoid rdkit to change the SMILES accodingly to sanitization rules

Additionally, you can try to use the implicit smiles syntax as below. I tested your smiles with random sequence and the output worked for me

• ligand:
  id: B
  smiles: '[CH3][CH2][CH2][CH2][CH2][CH2][CH2]N'