Using PMX for ΔΔG on a glycosylated Fc (P238D back-mutation; PDB 3WJJ): viable workflows?

[EidRashed]

Hi all,

I’d like to use PMX to compute ΔΔG of binding for an Fc that is N-glycosylated (structure based on PDB: 3WJJ) by reversing the mutation P238D. I know PMX supports Amber and CHARMM family protein force fields out of the box, but I’m unclear how best to proceed when linked glycans are present.

Concretely:

Amber/GLYCAM → GROMACS → PMX?
If I build the glycoprotein in tleap with Amber+GLYCAM, then convert to GROMACS formats (e.g., via ACPYPE or similar), is that topology/coordinate set acceptable input for pmx mutate? Key requirement: the covalent N-glycosidic bond (Asn–GlcNAc) must survive the conversion and be represented as a bonded connection in GROMACS. Has anyone run PMX successfully on such converted glycoprotein systems? (Background: GLYCAM→GROMACS conversion with ACPYPE is documented, but I’m not sure about PMX compatibility and preserving cross-residue bonds. )

All-CHARMM36 route with pdb2gmx + PMX CHARMM-mut ff?
Alternatively, prepare the glycoprotein entirely with CHARMM36 (jul2022) carbohydrate parameters in GROMACS (i.e., using carb.rtp/carb.r2b so mid-chain vs terminal sugar names map correctly), then use PMX’s CHARMM mutate workflow. Is there a recommended recipe (naming, r2b mappings for mid-chain GlcNAc, Man, Gal, Fuc, specbond.dat for linkages, etc.) known to work end-to-end with PMX? (GROMACS handles CHARMM36 glycans via carb.rtp + carb.r2b; correct residue naming and r2b mapping are critical. )

Any proven alternatives?
If the above aren’t practical, are there known workarounds (e.g., splitting protein/glycans, then merging with manual bonds in topology; or specific PMX mapping files for common N-glycans) that people have used successfully for ΔΔG with glycoproteins?

Motivation / constraints:
We prefer to avoid web services that require uploading proprietary coordinates (e.g., CHARMM-GUI), and we’re fine with some manual editing if that’s the current state-of-the-art. (We know CHARMM-GUI’s Glycan Reader/Modeler can produce GROMACS-ready glycoproteins; just checking for an offline/PMX-friendly path. )

Any pointers, example repos, or do/don’t tips (especially around PMX’s mutate step with carbohydrate-bearing topologies) would be hugely appreciated!

Thanks!

[wilsonjcarter]

Hi EidRashed,

As you alluded to, pmx mutations rely on running pdb2gmx on the mutant structure, which in turn requires a .ff directory containing the relevant glycosylated .rtp blocks (as mentioned in a previous response). The best approach is to generate a modified amino acid, in this case asparagine bound to the sugar chain.

I am not familiar enough with Amber, but here is a CHARMM-based approach for creating such a glycosylated amino acid. A similar strategy should be possible with Amber.

---

# 1. Generate tripeptide in CHARMM-GUI from an N-glycosylated tripeptide 
#    e.g., NH3+(Tyr-GlycoAsn-Ser)COO-; in this case extracted from PDB 3WJJ.
#    (This produces PROA.itp and a hydrogenated structure)

# 2. Parse the ITP to extract the modified residue and add to merged.rtp
python parse.py -itp PROA.itp

# 3. Extract/generate a glyco-modified asparagine residue (glyco_fragment.pdb) 
#    from the CHARMM-GUI structure

# 4. Run pdb2gmx on the non-glycosylated structure
gmx pdb2gmx -f 3WJJ_no_glyco.pdb -ff charmm36m-mut -ter -ignh -o gmx.pdb

# 5. Add the glycosylated residue to the structure
python dophos.py --input gmx.pdb \
                 --output gmx_glyco.pdb \
                 --target A:297 \
                 --fragment glyco_fragment.pdb \
                 --resname GAS

# 6. Run pdb2gmx again with the updated structure
gmx pdb2gmx -f gmx_glyco.pdb -ff charmm36m-mut -ter -o gmx_glyco2.pdb

parse.py
dophos.py
PROA.itp
glyco_fragment.pdb