Detection Metrics
Detection Metrics
Introduction
Oxonium Browser provides three primary metrics to assess the strength of a sugar oxonium ion match. Understanding these metrics and how to apply them is crucial for distinguishing genuine sugar signals from random mass peak matches.
Keep in mind that Oxonium Browser generates a list of oxonium ion candidates based on accurate mass, abundance values, and the occurrence of water loss peaks. The provided default parameters have been established based on a range of reference samples and serve as a good starting point.
Key Metrics
1. Spectral Counts
Definition: The number of MS2 spectra where a specific oxonium ion was detected above the intensity threshold.
Guidelines:
- Counts > 20 (default): reliable detection in most experiments
- Counts 10–20: moderate confidence, review alongside other metrics
- Counts < 10: low confidence, may represent rare or weak signals
Higher counts generally indicate more reliable detection, but some sugars may naturally yield lower counts depending on their abundance and fragmentation efficiency.
2. Spectral Intensity
Definition: The normalized intensity of the oxonium ion relative to total spectrum intensity, averaged across all detected spectra. Calculated as: (average intensity of the diagnostic mass pair / total spectrum intensity) × 100.
Guidelines:
- Default threshold: 0.2%
- Typical range for genuine oxonium ions: 0.2–6%
- Values as low as 0.05% have been found valid in some reference samples
Higher intensities suggest more confident signals. Low intensities may indicate weak detection or background noise.
3. Spectral Presence
Definition: The percentage of all MS2 spectra (excluding SAGE-identified peptides) where the oxonium ion was detected above the intensity threshold.
Guidelines:
- Default threshold: 0.02%
- Presence >1% often indicates abundant protein glycosylation
- Presence <0.02% is commonly associated with random background matches
This metric normalizes for dataset size, making it useful for comparing results across experiments with different numbers of spectra.
Using Test Masses as Negative Controls
The sugar database includes Ox_test_ entries — random masses serving as built-in negative controls. These are essential for determining appropriate thresholds. Any detections of test masses represent random chance matches, providing a direct estimate of the false positive rate at the current settings.
Test masses are generated within the sugar oxonium ion mass range (100–400 Da) with restricted mass defects similar to actual sugar fragments. For more details on how test masses are constructed, see Sugar Database.
Threshold Optimization Workflow
1. Start with default thresholds (counts: 20, intensity: 0.2%, presence: 0.02%)
2. Observe test mass behavior — note the distribution of test masses relative to real hits.
3. Adjust count threshold first — increase until most test masses disappear from the match table.
4. Fine-tune with intensity — if test masses remain, raise the intensity threshold.
5. Use presence for final refinement — adjust last, useful for normalizing across dataset sizes.
6. Verify across visualizations:
- Check the match table for test masses appearing in sugar groups
- Examine RT profiles — false positives often appear only at run boundaries
- Look at co-occurrence — genuine sugars from the same glycan should cluster together
Interpreting the Match Table
The match table groups oxonium ions into ±18 Da water loss families. A group containing an intact oxonium ion, its −H₂O fragment, and its −2H₂O fragment is strong evidence of a genuine sugar For a full description of the table, see Dashboard Guide.
Interpreting the Co-occurrence Plot
The clustered co-occurrence heatmap shows which selected oxonium ions appear together in the same spectra. The dendrogram groups ions by Jaccard similarity of their scan profiles. For details on how the plot is constructed, see Dashboard Guide.
Key patterns to look for:
- High co-occurrence (dark cells) between ions suggests they originate from the same glycan structure
- Low or zero co-occurrence suggests ions come from different glycopeptides or different proteins
- Tight dendrogram clusters indicate ions with very similar scan profiles — strong evidence they belong to the same glycan
- Test masses should show minimal co-occurrence with real ions — if they co-occur strongly, thresholds may need adjustment
Interpretation Strategies
Biological Context Matters
Consider your sample source when evaluating results. Oxonium ions from mammalian samples typically produce higher metrics across the board. Bacterial sugars may yield lower values but can still be genuine.
Related Sugar Series
Sugars belonging to the same glycan often show similar scan distribution patterns. Use the retention time profiles to check whether candidate sugars co-elute, and the co-occurrence plot to confirm they appear in the same spectra.
Common Pitfalls
Setting thresholds too low: Many false positives appear; test masses mix with genuine hits in the table.
Setting thresholds too high: Genuine signals from rare glycans or low-abundance sugars are missed.
Ignoring test mass performance: Test masses are the best guide to appropriate thresholds. If they persist after adjustment, interpret results with caution.
Ignoring retention time profiles: False positives often cluster at the beginning or end of the chromatographic run. Always check RT profiles for suspicious patterns.
