glycobiology

name	glycobiology
description	Glycosylation site prediction and glycobiology analysis. N-glycosylation motif finding, O-glycosylation hotspot prediction, glycan structure resources. Lightweight, pure Python. For protein function queries use uniprot-database; for structure analysis use alphafold-database.
category	biology
license	MIT license
metadata	{"skill-author":"InkVell Inc."}

Glycobiology: Glycosylation Analysis

Overview

Glycobiology provides lightweight computational tools for predicting and analyzing glycosylation sites in protein sequences. This skill covers N-glycosylation sequon motif finding (N-X-S/T where X is not P), O-glycosylation hotspot prediction using a sliding window serine/threonine density heuristic, glycan structure tool references, and combined glycoprotein analysis with domain mapping. All analyses use pure Python with minimal dependencies (Biopython for sequence I/O, regex for pattern matching).

When to Use This Skill

• Predicting N-glycosylation sites from protein sequences
• Identifying O-glycosylation hotspot regions
• Planning glycosylation site mutagenesis experiments
• Comparing predicted vs experimentally determined glycosylation sites
• Annotating glycosylation in biotherapeutic protein design
• Surveying glycan analysis tools for downstream structural studies

Related Skills: For protein function and existing glycosylation annotations use uniprot-database. For protein 3D structure and site accessibility use alphafold-database. For sequence manipulation use biopython.

Installation

uv pip install biopython numpy

No additional dependencies required — this skill uses pure Python.

Quick Start

import re

def find_n_glycosylation_sites(sequence):
    """Find N-X-S/T sequons where X != P."""
    sites = []
    for i in range(len(sequence) - 2):
        if sequence[i] == 'N' and sequence[i+1] != 'P' and sequence[i+2] in ('S', 'T'):
            sites.append({
                'position': i + 1,  # 1-based
                'motif': sequence[i:i+3],
                'context': sequence[max(0,i-3):i+6]
            })
    return sites

# Example: human EPO
epo_seq = "MGVHECPAWLWLLLSLLSLPLGLPVLGAPPRLICDSRVLERYLLEAKEAENITTGCAEHCSLNENITVPDTKVNFYAWKRMEVGQQAVEVWQGLALLSEAVLRGQALLVNSSQPWEPLQLHVDKAVSGLRSLTTLLRALGAQKEAISPPDAASAAPLRTITADTFRKLFRVYSNFLRGKLKLYTGEACRTGDR"

sites = find_n_glycosylation_sites(epo_seq)
print(f"N-glycosylation sites: {len(sites)}")
for s in sites:
    print(f"  Position {s['position']}: {s['motif']} (context: ...{s['context']}...)")

Core Capabilities

1. N-Glycosylation Motif Finding

Identify N-X-S/T sequons in protein sequences.

import re
from Bio import SeqIO

def find_n_glycosylation_sites(sequence, exclude_proline=True):
    """Find N-linked glycosylation sequons (N-X-S/T).

    Args:
        sequence: protein sequence string
        exclude_proline: if True, exclude N-P-S/T motifs (standard rule)
    """
    sites = []
    seq = str(sequence).upper()

    for i in range(len(seq) - 2):
        if seq[i] != 'N':
            continue

        x_residue = seq[i + 1]
        st_residue = seq[i + 2]

        # Standard rule: X != P
        if exclude_proline and x_residue == 'P':
            continue

        if st_residue not in ('S', 'T'):
            continue

        # Context (5 residues each side)
        context_start = max(0, i - 5)
        context_end = min(len(seq), i + 8)
        context = seq[context_start:context_end]

        sites.append({
            'position': i + 1,  # 1-based position
            'residue': seq[i],
            'motif': seq[i:i+3],
            'x_residue': x_residue,
            'acceptor': st_residue,
            : context,
            : context_start + 
        })

     sites

 ():
    
    results = []
     record  SeqIO.parse(fasta_path, ):
        sites = find_n_glycosylation_sites((record.seq))
        results.append({
            : record.,
            : record.description,
            : (record.seq),
            : (sites),
            : sites,
            : (sites) / (record.seq) *   
        })

     pandas  pd
    df = pd.DataFrame([{k: v  k, v  r.items()  k != }  r  results])
    ()
    ()
    ()
     results


sites = find_n_glycosylation_sites()
 s  sites:
    ()

2. O-Glycosylation Hotspot Prediction

Identify regions enriched in serine/threonine that may be O-glycosylated.

import numpy as np

def predict_o_glycosylation_hotspots(sequence, window_size=20, threshold=0.4,
                                       exclude_near_proline=True):
    """Predict O-glycosylation hotspots using S/T density heuristic.

    O-glycosylation preferentially occurs in S/T-rich regions, often near
    proline residues. This heuristic identifies candidate regions.

    Args:
        sequence: protein sequence string
        window_size: sliding window size
        threshold: minimum S/T fraction to call a hotspot
        exclude_near_proline: if True, downweight S/T not near P
    """
    seq = str(sequence).upper()
    n = len(seq)

    if n < window_size:
        return []

    # Calculate S/T density in sliding windows
    densities = []
    for i in range(n - window_size + 1):
        window = seq[i:i + window_size]
        st_count = window.count('S') + window.count('T')
        density = st_count / window_size
        densities.append(density)

    densities = np.array(densities)

    # Find hotspot regions
    hotspots = []
    in_hotspot = False
    start = 0

    for i, d in enumerate(densities):
        if d >= threshold and not in_hotspot:
            start = i
            in_hotspot = True
        elif d < threshold and in_hotspot:
            hotspots.append({
                'start': start + 1,  # 1-based
                'end': i + window_size,
                'length': i + window_size - start,
                : (densities[start:i].()),
                : (densities[start:i].mean()),
                : seq[start:i + window_size]
            })
            in_hotspot = 

     in_hotspot:
        hotspots.append({
            : start + ,
            : n,
            : n - start,
            : (densities[start:].()),
            : (densities[start:].mean()),
            : seq[start:]
        })

    
    all_st_sites = []
     hs  hotspots:
        region_start = hs[] - 
        region_seq = hs[]
         j, aa  (region_seq):
             aa  (, ):
                pos = region_start + j + 
                
                has_nearby_P = 
                 k  (-, ):
                    check_pos = region_start + j + k
                      <= check_pos < n  seq[check_pos] == :
                        has_nearby_P = 
                        

                all_st_sites.append({
                    : pos,
                    : aa,
                    : ,
                    : has_nearby_P
                })

    ()
     hs  hotspots:
        (
              )
    ()

     hotspots, all_st_sites

3. Glycan Structure Resources

Curated list of tools for glycan structure analysis and prediction.

GLYCAN_TOOLS = {
    'NetNGlyc': {
        'url': 'https://services.healthtech.dtu.dk/services/NetNGlyc-1.0/',
        'description': 'Neural network prediction of N-glycosylation sites',
        'input': 'Protein sequence (FASTA)',
        'output': 'N-glycosylation probability per Asn',
        'note': 'More accurate than motif-only scanning; considers sequence context'
    },
    'NetOGlyc': {
        'url': 'https://services.healthtech.dtu.dk/services/NetOGlyc-4.0/',
        'description': 'Neural network prediction of mucin-type O-glycosylation',
        'input': 'Protein sequence (FASTA)',
        'output': 'O-glycosylation probability per Ser/Thr',
        'note': 'Version 4.0 trained on O-GalNAc glycoproteomics data'
    },
    'GlycoWorkbench': {
        'url': 'https://code.google.com/archive/p/glycoworkbench/',
        'description': 'Glycan structure drawing and MS annotation',
        'input': 'Mass spectrometry data',
        'output': 'Glycan structure assignments'
    },
    'GLYCAM-Web': {
        'url': 'https://glycam.org/',
        'description': 'Glycan 3D structure building and MD simulation',
        'input': 'Glycan sequence (IUPAC condensed)',
        'output': '3D coordinates (PDB), AMBER parameters'
    },
    'Glycoshield-MD': {
        : ,
        : ,
        : ,
        : 
    },
    : {
        : ,
        : ,
        : ,
        : 
    }
}

 ():
    
     name, info  GLYCAN_TOOLS.items():
        ()
        ()
        ()
           info:
            ()

4. Glycoprotein Analysis

Combined N/O glycosylation analysis with domain mapping.

def analyze_glycoprotein(sequence, protein_name='protein',
                          uniprot_sites=None):
    """Comprehensive glycosylation analysis.

    Args:
        sequence: protein sequence
        protein_name: name for reporting
        uniprot_sites: dict of known sites from UniProt
                       {'N_glyc': [positions], 'O_glyc': [positions]}
    """
    seq = str(sequence).upper()
    print(f"=== Glycoprotein Analysis: {protein_name} ===")
    print(f"Length: {len(seq)} residues")

    # N-glycosylation
    n_sites = find_n_glycosylation_sites(seq)
    print(f"\nN-glycosylation sites (sequon motif): {len(n_sites)}")
    for s in n_sites:
        print(f"  N{s['position']}: {s['motif']}")

    # O-glycosylation hotspots
    o_hotspots, o_sites = predict_o_glycosylation_hotspots(seq)

    # Comparison with UniProt annotations
    if uniprot_sites:
        known_n = set(uniprot_sites.get('N_glyc', []))
        predicted_n = set(s['position'] for s in n_sites)

        tp = known_n & predicted_n
        fp = predicted_n - known_n
        fn = known_n - predicted_n

        print(f"\nComparison with UniProt annotations:")
        print(f"  True positives (predicted & annotated): {len(tp)}")
        ()
        ()

         fn:
            ()
             pos  fn:
                context = seq[(,pos-):pos+]
                ()

    
    ()
    ()
    ()
    ()
    ()

     {
        : n_sites,
        : o_hotspots,
        : o_sites
    }

Typical Workflows

Workflow 1: Predict All Glycosylation Sites in a Protein

sequence = "MANLQTSPNGTLKNVTSITDANLTQSPNNFTLRNITSANDTVTQSTPNVTQLNVSQSTPNVTQLNVSQST"
results = analyze_glycoprotein(sequence, protein_name='MyProtein')

Workflow 2: Compare Predicted vs UniProt-Annotated Sites

sequence = "MGVHECPAWLWLLLSLLSLPLGLPVLGAPPRLICDSRVLERYLLEAKEAENITTGCAEHCSLNENITVPDT"
known = {'N_glyc': [24, 38, 83], 'O_glyc': []}  # Known from UniProt
results = analyze_glycoprotein(sequence, protein_name='EPO', uniprot_sites=known)

Best Practices

1. N-glycosylation prediction — sequon motif (N-X-S/T, X!=P) is necessary but not sufficient; not all sequons are glycosylated; use NetNGlyc for probability-based prediction
2. O-glycosylation — much harder to predict than N-glyc; the S/T density heuristic identifies candidate regions but has high false positive rate; use NetOGlyc for better accuracy
3. Validate with UniProt — always cross-reference predictions with experimentally verified glycosylation annotations in UniProt
4. Signal peptide — glycosylation occurs in the ER/Golgi; ensure the protein has a signal peptide (positions are relative to mature protein in databases)
5. Structural context — glycosylation requires surface accessibility; use AlphaFold to check if predicted sites are surface-exposed
6. Biotherapeutics — for antibody/protein drug design, N-glycosylation in Fc regions (N297 in IgG) is critical for effector function

Troubleshooting

Problem: Sequon search finds sites in cytoplasmic proteins Solution: N-glycosylation only occurs in secretory pathway proteins. Filter predictions by subcellular localization (check UniProt).

Problem: Known glycosylation site not detected Solution: Some rare N-glycosylation occurs at non-canonical motifs (N-X-C). Check if the site has N-P-S/T which is excluded by default.

Problem: Too many O-glycosylation hotspots Solution: Increase the density threshold (e.g., 0.5). Focus on regions that are both S/T-rich and in extracellular domains.

Resources

• UniProt Glycosylation Annotations
• NetNGlyc Server
• NetOGlyc Server
• GlyConnect Database
• Essentials of Glycobiology (NCBI)