macrocycle

[mmagithub]

Hi,

I was wondering if CREM can be modified or adapted to generate linkers between two separate molecules, specifically at predefined stitching sites, to form a closed macrocycle.

Do you have any suggestions on how this might be achieved?

Thanks,
M

[DrrDom]

Hi,

this is an interesting question. There is no such functionality in crem. However, I think this can be achieved.

1. Set atom properties for atoms you want to link. This will create identifiers which will not be lost during next steps.
2. Break a bond in an input molecule and label atoms formed this bond with specific labels or again set atom properties as identifiers. Store two parts as individual molecules.
3. Run link_mols submitting these two molecules and providing corresponding atom ids retrieved by searching of a particular atom field (atoms ids will change after splitting a molecule, therefore we need those additional identifiers).
4. Create a bond between two atoms labeled at the step 2.

Instead of setting atom properties, you may set isotope labels and then remove them.

The only downside of this procedure, which I see, is that if a bond you cut is close to linking atoms (distance is less than a chosen radius), this will affect fragments selected for linking, because this will change linking atom context.

I think this may be a useful feature to implement as a wrapper around link_mols. You may create such a function and contribute to the module.

Actually the same approach may be used to create new ring systems in general and overcome the main limitation of the approach - inability to create new rings. However, this will require some changes in the core code. When fragments will be created we have to cut rings, label atoms at a broken bond with specific labels (need to think about suitable and universal labeling), fragment a molecule as usual and create a fragment database,. Thus we will have a specific atom labels, which will be recognized as a valid chemical context in the generation functions.

[mmagithub]

I will think about that, thanks for your suggestions.
At the moment, I am using the linkers from ProtacsDB, I think this can be adopted to work in CREM as well, and I do not think that having the context radius constrain/functionality is mandatory here, I believe the linkers already have all potential combinations that can be post-filtered.
http://cadd.zju.edu.cn/protacdb/downloads

The linkers are already annotated by R1/R2 from the stitching sites, and yes, I am using something similar to what you described to annotate those sites on the fragments, using flourine placeholders that will be removed upon stitching the linkers to the amide/reverse amide that these flourines are attached to. This is a very specific problem though, puting something similar into CREM may require ability to generalise to more scenarios.

[DrrDom]

Yes, your case is rather specific. However, it would be reasonable to think about a general solution. Thanks for initiating this discussion