Native glycoprotein support in the OpenMMDL Setup AMBER path

[MarvinTaterra]

Native support for N-/O-glycosylated proteins in the OpenMMDL Setup AMBER path. A glycoprotein PDB (protein with covalently attached glycans) can now be taken to a runnable MD system directly, without an external preparation step.

Motivation

Neither existing path handles glycoproteins today:

• PDB path — OpenMM's bundled force-field XMLs carry no GLYCAM templates, so createSystem fails with No template found for residue NAG.
• AMBER path — only one receptor force field is sourced, and no bond statements are emitted, so the protein–glycan linkages are lost.

Changes

New openmmdl/openmmdl_setup/glycoprotein.py — glycan detection and
GLYCAM renaming:

• Detects sugar residues by PDB Chemical Component Dictionary name
  (NAG, NDG, BMA, MAN, FUC, FUL).
• Finds protein–glycan and glycan–glycan linkages from CONECT
  records, with a distance-based fallback when they are absent.
• Maps residues and atoms to GLYCAM-06j-1 conventions (e.g.
  NAG → 0YB/4YB, glycosylated ASN → NLN).

glycoprotRcp receptor type in the AMBER script generation:

• Sources both the protein force field and the GLYCAM carbohydrate
  force field.
• Emits an explicit bond statement for every glycosidic linkage so
  they survive into the generated topology.
• Cleans the GLYCAM-renamed structure with pdb4amber --no-reduce.

Glycan-aware PBC imaging in post_md_conversions.py — keeps glycans
in the same periodic image as the protein residue they are bonded to
during trajectory re-imaging.

UI — a "Glycoprotein" option in the Amber Setup receptor tab.

Scope and limitations

Supports the common N-glycan core sugars listed above. The GLYCAM
residue codes, linkage codes and atom-name maps were verified against
the GLYCAM-06j-1 prep/lib files in
GLYCAM-Web/gmml. Sugars or
linkage patterns outside the verified set raise a descriptive
NotImplementedError rather than producing an incorrect topology —
extending coverage is a matter of adding verified table entries.

PDB-path (OpenMM XML) glycoprotein support would require custom
residue templates and is left as future work.

Testing

32 new tests under openmmdl/tests/ — unit tests for detection and
renaming, an integration test for the generated AMBER script, and an
analysis test for the imaging helper. The full pipeline was verified
end-to-end on a four-chain N-glycosylated factor XIa system:
detection → tleap topology generation → OpenMM minimization and MD ran