Wrong sugammadex conformation

[Chebuu]

Sugammadex functional groups are oriented in a sort of "trans" conformation (180° dihedral about the glycosidic bond). This conformation disagrees with both experimental data and previously reported MD simulations https://doi.org/10.1016/j.molliq.2018.06.033

I think the problem is that RDKit ends up minimizing a 2D (planar) structure with strong non-bonded interactions between the sugar hydroxyl groups.

Ultimately, this issue highlights an important problem when assembling structures from 2D gaphs, so I need a reliable/consistent solution before I start generating HTS libraries with stk.

A constrained embedding i.e. rdkit.Chem.AllChem.ConstrainedEmbed might be a good option, but I'd rather avoid it.

[Chebuu]

On second thought, Avogadro optimizes all my other 2D cyclodextrin conformers as expected when no "guest" molecule is present. So I'm guessing RDKit will generate the correct conformer too. If it does, then I think I can just initialize the sugammadex building block from an RDKit molecule and use it to construct the host-guest complex, because I'm pretty sure BuildingBlock.init_from_rdkit_mol preserves atom positions.

[Chebuu]

See this section from the RDKit Cookbook about improved torsions for macrocycles with ETKDGv3:

In 2020, we devised some improvements to the ETKDG method for sampling small rings and macrocycles 5.

# this includes addtional small ring torsion potentials
params = AllChem.srETKDGv3()

# this includes additional macrocycle ring torsion potentials and macrocycle-specific handles
params = AllChem.ETKDGv3()

# to use the two in conjunction, do:
params = AllChem.ETKDGv3()
params.useSmallRingTorsions = True

# a macrocycle attached to a small ring
mol = Chem.MolFromSmiles("C(OC(CCCCCCC(OCCSC(CCCCCC1)=O)=O)OCCSC1=O)N1CCOCC1")
mol = Chem.AddHs(mol)
AllChem.EmbedMultipleConfs(mol, numConfs = 3 , params = params)