Cis amides in cyclic peptides

[isohelio]

Hi,

I'm using pyPept to build simple peptide macrocycles. I'm finding the BILN notation extremely useful for this and the pyPept interface is very straightforward to use.

Unfortunately, there seems to be a problem with the geometry of peptide bonds in cyclic peptides. If I build a simple linear deca-alanine then the geometry is all sensible, entirely trans peptide bond (N-H points away from C=O).

run_pyPept --biln A-A-A-A-A-A-A-A-A-A

If I cyclise that structure then essentially all of the peptide bonds become cis (N-H points in same direction as C=O).

run_pyPept --biln 'A(1,1)-A-A-A-A-A-A-A-A-A(1,2)'

I think non-proline peptide bonds are nearly always trans even in relatively small cyclic peptides.
e.g. as seen in PDB structure 7zkr.

Is this a known problem, or a misunderstanding on my part?

Many thanks,
Mikie

[isohelio]

Attaching pdb file output for the two structures.

pept.zip

[amin-sagar]

Hello.
I have the same observation. I tried to generate a bicyclic peptide as mentioned in #9 and all the peptide bonds are in cis conformation.

Best,
Amin.

[foxth]

Ah, that's a problem of the underlying rdkit routine to produce a 3D structure. We simply pass our molecule object to rdkit's ETKDGv3 coordinate generator. And as this is a distance geometry procedure, anything might come out of this. For small ring sizes, cis-amides are the only way to get the ring closed, so Im not sure there is a general way to avoid this. Maybe playing with some of the options of ETKDGv3 might help, but I havent tried these.
This is even a problem when generating 2D structures (again, here we rely on rdkits rdDepictor module): I have generated cyclic poly-Ala with different ring sizes, and the larger they get the better rdkit is in producing trans-amides.

But even this doesnt always help you: for cAla10, the 2D structure looks perfectly fine, all trans, but when I load that into moe and do a minimization to obtain a 3D conformer, I end up with one cis bond in the peptide.

So no easy solution here, Im afraid. Maybe posting the problem to the rdkit list - they might have a solution to this (I believe there are ways to set some constraints for the DG algorithm, so one would be a minimum distance between the NH and the carbonyl oxygen for all the amide bonds, but I have never done this). I just skimmed through the rdkit documentation - there is a flag to enforce trans-amides (https://www.rdkit.org/docs/source/rdkit.Chem.rdDistGeom.html -> forceTransAmides); this could be the way to go...

Best,
Th.

[isohelio]

Hi,

thanks for the assessment. I was just coming to the same conclusion that the problem was arising in RDKit.

I think trans peptides are almost universal even in small peptide macrocycles. Here is one from PDB 7tb1.

I had not seen that setting for trans peptides, will see if I can activate it.

Regards,
Mike

[isohelio]

Hi,

Ok, I modified conformer.py after line ~317 as shown below

            parameters.useRandomCoords = True
            print("settings forceTransAmides")
            parameters.forceTransAmides = True
            AllChem.EmbedMolecule(romol, parameters)

Running cyclic alanine reports

settings forceTransAmides
15:02:29   INFO:File generated: cyclic_deca_alanine.png.
15:02:29   INFO:File generated: cyclic_deca_alanine.pdb.

But alas output is still all cis.

Reading rdkit/rdkit#3794 seems to suggest that it is the default.

I will report the problem at RDKit, and see if there is any possibility for a solution.

Thanks
Mike