Mass to charge ratio validation

mass-to-charge-ratio-validation

Summary

Excludes metabolomic features with implausible or instrument-artifact m/z values by filtering based on mass defect thresholds. This preprocessing step removes spurious signals that arise from instrument noise or acquisition artifacts before downstream statistical and identification analysis.

When to use

Apply this skill after loading MS-Dial feature tables (e.g., Urine_RP_NEG_norm.txt or Urine_RP_POS_norm.txt) and before sample-level filtering or imputation, whenever the feature abundance matrix contains m/z values acquired across multiple chromatographic runs or polarities. Use it if your dataset exhibits features with m/z decimal components that are biologically implausible or fall within known instrumental artifact ranges.

When NOT to use

• Input is already a curated feature table from a commercial or vendor-validated preprocessing pipeline that has already removed m/z artifacts.
• Your study uses only a single, well-characterized instrument run with known stable calibration; instrumental artifacts are negligible.
• You are analyzing targeted metabolomics data where features have been pre-selected by expert knowledge and validated against authentic standards; mass defect filtering may remove valid low-abundance metabolites.

Inputs

• MS-Dial feature table (text file: e.g., Urine_RP_NEG_norm.txt, Urine_RP_POS_norm.txt)
• margheRita data structure (after loading MS-Dial output)
• Feature abundance matrix with m/z and feature annotations

Outputs

• Filtered feature abundance matrix (features with artifact m/z values removed)
• Retained feature count and exclusion report (e.g., 548/604 features retained)
• Feature metadata with m/z validation flags

How to apply

The filtering() function in margheRita applies mass defect validation by removing features whose m/z decimal values fall within the range [4, 8] by default. This range captures common acquisition artifacts. Load the MS-Dial output as a margheRita data structure, then invoke the filtering function with the m/z filtering step enabled. The rationale is that true metabolites have m/z values distributed across the full decimal range; clustering within [4, 8] signals instrumental or chemical noise. On the Urine dataset, this step reduced features from 604 to 548 (56 features removed, ~9% of total), confirming artifact removal without excessive feature loss. Verify the threshold is appropriate for your instrument and polarity by inspecting the m/z distribution histogram before and after filtering.

Related tools

• margheRita (Provides the filtering() function that implements mass defect-based m/z validation on MS-Dial output feature tables) — https://github.com/emosca-cnr/margheRita
• MS-Dial (Peak picking software that generates the input feature tables (Urine_RP_NEG_norm.txt, Urine_RP_POS_norm.txt) subject to m/z validation)
• R (Runtime environment for executing margheRita filtering functions)

Examples

filtered_data <- filtering(urine_data, mass_defect_range = c(4, 8))

Evaluation signals

• Feature count before and after m/z filtering matches reported retention (e.g., 604 → 548 features on Urine dataset implies ~56 features removed).
• Histogram of m/z decimal values shows gap or depletion in the [4, 8] range post-filtering, confirming artifact removal.
• No features with m/z decimal values in [4, 8] remain in the output feature table.
• Sample count is unchanged (filtering is feature-level only; e.g., 243/243 samples retained on Urine dataset).
• Downstream statistical tests and metabolite identification show improved specificity or reduced false-positive annotations compared to unfiltered data.

Limitations

• The default m/z decimal range threshold [4, 8] is empirically determined from the Urine dataset and may not generalize to other instrument types, chromatographic methods, or polarities; validation on your own data is recommended.
• Mass defect filtering alone cannot distinguish between true low-abundance metabolites and instrument noise; it is most effective when combined with sample-level and feature-level missing-value filtering.
• Features with valid metabolite annotations that happen to fall within [4, 8] m/z decimal range risk false-positive exclusion; inspection of removed features against spectral libraries is prudent.
• The skill assumes m/z values are properly calibrated by MS-Dial; miscalibration will render this filtering step ineffective.

Evidence

• [other] removes features with m/z decimal values within the range [4, 8] by default: "removes features with m/z decimal values within the range [4, 8] by default, and (3) imputes remaining NA values"
• [other] On the Urine dataset, this retained 243/243 samples, 604/604 features after NA filtering, and 548/604 features after m/z filtering: "On the Urine dataset, this retained 243/243 samples, 604/604 features after NA filtering, and 548/604 features after m/z filtering."
• [intro] runs filters to exclude features/sample with many missing values, features with wrong m/z values: "runs filters to exclude features/sample with many missing values, features with wrong m/z values"
• [intro] filtering by mass defects, filtering by coefficient of variation (samples vs QCs) and probabilistic quotient normalization: "filtering by mass defects, filtering by coefficient of variation (samples vs QCs) and probabilistic quotient normalization"
• [intro] margheRita is intended to be used after having done a number of data acquisition steps through MS-Dial: "margheRita is intended to be used after having done a number of data acquisition steps through MS-Dial"