Analyze and engineer protein glycosylation. Scan sequences for canonical N-glycosylation sequons (N-X-S/T with X not proline, including overlapping sites), predict O-GalNAc hotspots, read glycan notation, and reach the curated external tooling (NetNGlyc, NetOGlyc, GlycoShield, GlycoWorkbench, GlyTouCan, GlyConnect). Use this skill for therapeutic antibody glycoengineering and afucosylation for ADCC, Fc glycan control, glycan shielding in vaccine immunogen design, sequon removal or insertion, and half-life engineering through sialylation. Also trigger on N-glycosylation, sequon, NXS/NXT, O-glycosylation, glycoform heterogeneity, afucosylation, high-mannose, GlyTouCan, or WURCS.
Analyze and engineer protein glycosylation. Scan sequences for canonical N-glycosylation sequons (N-X-S/T with X not proline, including overlapping sites), predict O-GalNAc hotspots, read glycan notation, and reach the curated external tooling (NetNGlyc, NetOGlyc, GlycoShield, GlycoWorkbench, GlyTouCan, GlyConnect). Use this skill for therapeutic antibody glycoengineering and afucosylation for ADCC, Fc glycan control, glycan shielding in vaccine immunogen design, sequon removal or insertion, and half-life engineering through sialylation. Also trigger on N-glycosylation, sequon, NXS/NXT, O-glycosylation, glycoform heterogeneity, afucosylation, high-mannose, GlyTouCan, or WURCS.
license
MIT
compatibility
Requires Python 3.10+. The bundled sequon and notation analysis is standard library only. Optional extras — pandas and requests for the database lookups, glycoshield for ensemble modelling. The external predictors (NetNGlyc, NetOGlyc) are DTU web services requiring manual submission and, for some, an academic licence; there is no public REST API.
allowed-tools
Read Write Edit Bash
metadata
{"version":"1.2","skill-author":"Kuan-lin Huang"}
Glycoengineering
Overview
Glycosylation is the most common and complex post-translational modification (PTM) of proteins, affecting over 50% of all human proteins. Glycans regulate protein folding, stability, immune recognition, receptor interactions, and pharmacokinetics of therapeutic proteins. Glycoengineering involves rational modification of glycosylation patterns for improved therapeutic efficacy, stability, or immune evasion.
Two major glycosylation types:
N-glycosylation: Attached to asparagine (N) in the sequon N-X-[S/T] where X ≠ Proline; occurs in the ER/Golgi
O-glycosylation: Attached to serine (S) or threonine (T); no strict consensus motif; primarily GalNAc initiation
When to Use This Skill
Use this skill when:
Antibody engineering: Optimize Fc glycosylation for enhanced ADCC, CDC, or reduced immunogenicity
Therapeutic protein design: Identify glycosylation sites that affect half-life, stability, or immunogenicity
Vaccine antigen design: Engineer glycan shields to focus immune responses on conserved epitopes
Biosimilar characterization: Compare glycan patterns between reference and biosimilar
Drug target analysis: Does glycosylation affect target engagement for a receptor?
Protein stability: N-glycans often stabilize proteins; identify sites for stabilizing mutations
N-Glycosylation Sequon Analysis
Scanning for N-Glycosylation Sites
N-glycosylation occurs at the sequon N-X-[S/T] where X ≠ Proline.
import re
from typing importList, Tupledeffind_n_glycosylation_sequons(sequence: str) -> List[dict]:
"""
Scan a protein sequence for canonical N-linked glycosylation sequons.
Motif: N-X-[S/T], where X ≠ Proline.
Args:
sequence: Single-letter amino acid sequence
Returns:
List of dicts with position (1-based), motif, and context
"""
seq = sequence.upper()
results = []
# Step by one, not by three. Sequons overlap: in NNTS both N1 (NNT) and
i ((seq) - ):
triplet = seq[i:i+]
triplet[] == triplet[] != triplet[] {, }:
results.append({
: i + ,
: triplet,
: seq[(, i-):i+],
: triplet[] ==
})
results
() -> :
sequons = find_n_glycosylation_sequons(sequence)
lines = []
lines.append()
lines.append()
sequons:
lines.append()
lines.append()
lines.append()
s sequons:
lines.append()
:
lines.append()
.join(lines)
fc_sequence =
(summarize_glycosylation_sites(fc_sequence, ))
# N2 (NTS) are glycosylation sites, and advancing past a match would
# silently drop the second one -- a missed liability, reported as clean.
for
in
range
len
2
3
if
0
'N'
and
1
'P'
and
2
in
'S'
'T'
'position'
1
# 1-based, the asparagine
'motif'
'context'
max
0
3
6
# ±3 residue context
'sequon_type'
'NXS'
if
2
'S'
else
'NXT'
return
def
summarize_glycosylation_sites
sequence: str, protein_name: str = ""
str
"""Generate a research log summary of N-glycosylation sites."""
defeliminate_glycosite(sequence: str, position: int, replacement: str = "Q") -> str:
"""
Eliminate an N-glycosylation site by substituting Asn → Gln (conservative).
Args:
sequence: Protein sequence
position: 1-based position of the Asn to mutate
replacement: Amino acid to substitute (default Q = Gln; similar size, not glycosylated)
Returns:
Mutated sequence
"""
seq = list(sequence.upper())
idx = position - 1assert seq[idx] == 'N', f"Position {position} is '{seq[idx]}', not 'N'"
seq[idx] = replacement.upper()
return''.join(seq)
defadd_glycosite(sequence: str, position: int, flanking_context: str = "S") -> str:
"""
Introduce an N-glycosylation site by mutating a residue to Asn,
and ensuring X ≠ Pro and +2 = S/T.
Args:
position: 1-based position to introduce Asn
flanking_context: 'S' or 'T' at position+2 (if modification needed)
"""
seq = list(sequence.upper())
idx = position - 1# Mutate to Asn
seq[idx] = 'N'# Ensure X+1 != Pro (mutate to Ala if needed)if idx + 1 < len(seq) and seq[idx + 1] == 'P':
seq[idx + 1] = 'A'# Ensure X+2 = S or Tif idx + 2 < len(seq) and seq[idx + 2] notin ('S', 'T'):
seq[idx + 2] = flanking_context
return''.join(seq)
O-Glycosylation Analysis
Heuristic O-Glycosylation Hotspot Prediction
defpredict_o_glycosylation_hotspots(
sequence: str,
window: int = 7,
min_st_fraction: float = 0.4,
disallow_proline_next: bool = True) -> List[dict]:
"""
Heuristic O-glycosylation hotspot scoring based on local S/T density.
Not a substitute for NetOGlyc; use as fast baseline.
Rules:
- O-GalNAc glycosylation clusters on Ser/Thr-rich segments
- Flag Ser/Thr residues in windows enriched for S/T
- Avoid S/T immediately followed by Pro (TP/SP motifs inhibit GalNAc-T)
Args:
window: Odd window size for local S/T density
min_st_fraction: Minimum fraction of S/T in window to flag site
"""if window % 2 == 0:
window = 7
seq = sequence.upper()
half = window // 2
candidates = []
for i, aa inenumerate(seq):
if aa notin ('S', 'T'):
continueif disallow_proline_next and i + 1 < len(seq) and seq[i+1] == 'P':
continue
start = max(0, i - half)
end = min(len(seq), i + half + 1)
segment = seq[start:end]
st_count = sum(1for c in segment if c in ('S', 'T'))
frac = st_count / len(segment)
if frac >= min_st_fraction:
candidates.append({
'position': i + 1,
'residue': aa,
'st_fraction': round(frac, 3),
'window': f"{start+1}-{end}",
'segment': segment
})
return candidates
External Glycoengineering Tools
1. NetOGlyc 4.0 (O-glycosylation prediction)
Web service for high-accuracy O-GalNAc site prediction:
Output: Per-residue O-glycosylation probability scores
Method: Neural network trained on experimentally verified O-GalNAc sites
import requests
defsubmit_netoglycv4(fasta_sequence: str) -> str:
"""
Submit sequence to NetOGlyc 4.0 web service.
Returns the job URL for result retrieval.
Note: This uses the DTU Health Tech web service. Results take ~1-5 min.
"""
url = "https://services.healthtech.dtu.dk/cgi-bin/webface2.cgi"# NetOGlyc submission (parameters may vary with web service version)# Recommend using the web interface directly for most use casesprint("Submit sequence at: https://services.healthtech.dtu.dk/services/NetOGlyc-4.0/")
return url
# Also: NetNGlyc for N-glycosylation prediction# URL: https://services.healthtech.dtu.dk/services/NetNGlyc-1.0/
2. GlycoShield-MD (Glycan Shielding Analysis)
GlycoShield-MD analyzes how glycans shield protein surfaces during MD simulations:
Start with NetNGlyc/NetOGlyc for computational prediction before experimental validation
Verify with mass spectrometry: Glycoproteomics (Byonic, Mascot) for site-specific glycan profiling
Consider site context: Not all predicted sequons are actually glycosylated (accessibility, cell type, protein conformation)
For antibodies: Fc N297 glycan is critical — always characterize this site first
Use GlyConnect to check if your protein of interest has experimentally verified glycosylation data
Composing with the rest of the bundle
antibody-engineering → before: an N-glycosylation sequon inside a CDR is a developability
liability, not a feature. Its liability scanner flags them with region weighting; come here to
decide whether to remove the sequon or keep it.
esm / protein-binder-design → after: scan any designed or generated sequence. Design tools
introduce sequons without noticing, and the glycan then blocks the interface you designed.
uniprot-rcsb → before: the canonical sequence, plus UniProt's own CARBOHYD annotations, which
tell you which sequons are actually occupied rather than merely present.
immunogenicity → alongside: glycan shielding and T-cell epitope exposure are the same surface
argued from two directions.
molecular-dynamics → after: GlycoShield-style ensembles are what turn a sequon list into an
actual picture of what the glycan covers.
A sequon is necessary, not sufficient. Occupancy depends on local structure and expression
system; roughly a third of predicted sequons are unoccupied. Report predicted sites as predicted.
Resources
Read references/glycan_databases.md when you need to look a
structure up rather than scan a sequence — GlyTouCan accessions, GlyConnect site-specific data,
UniCarbKB, the notation formats (GlycoCT, WURCS, IUPAC) and worked lookup code.
Review: Apweiler R, Hermjakob H, Sharon N (1999) On the frequency of protein glycosylation,
as deduced from analysis of the SWISS-PROT database. Biochim Biophys Acta. PMID: 10580125
Therapeutic glycoengineering review: Jefferis R (2009) Glycosylation as a strategy to improve
antibody-based therapeutics. Nature Reviews Drug Discovery. PMID: 19247305
