Hi Hari, On 16/09/10 19:57, hari jayaram wrote:
I am using the SMILES input based , ligand coordinate and restraint generation inside coot , which I understand calls libcheck. For some of these ligands , the stereochemistry around parts of the molecule are "wrong" : in that the SMILES string has solid bonds instead of wedges. I am using RDKIT the cheminformatics package to visualize these molecules.
I am slightly confused about what you are doing. Just so that we are on the same page: Libcheck produces a cif dictionary and a pdb file for the ligand (give a SMILES string).
How are you visualising these things with RDKit? (that is presumably a 2D representation, right?)
For these SMILES strings the model generated by libcheck inside coot is very mangled. Such that rings circle around and clash with each other for example.
I have had mixed results with multi-ring systems with libcheck.
For the present moment I am manually editing the molecule to correct the stereochemistry- generate the SMILES string externally and then have libcheck do the right thing.
OK, I don't understand this. What other way is there to generate SMILES strings? (other than "external") I mean.
Is there a way to avoid this second step.
possibly. You can generate a mol2 file from the SMILES, right? How about feeding that (or the SMILES string) to cprodrg?
(Not that cprodrg is entirely without flaws in multi-ring systems).
I would think that the default model generated my SMILES input inside coot should at-least not have rings forming knots with each other.
That is not an unreasonable desire. The version of Coot that you are using doesn't do anything with SMILES strings, other than hand them off to libcheck.
Also does anyone know if there is a way to flatten SMILES representations -i.e render them devoid of stereochemistry
You mean by this remove the @ and @@s right? Isn't that something that RDKit can do?
and then have libcheck do the most stable thing.
Hmm... I'm not sure that I follow that.
