mass-spectrometry-data-statistical-testing
Summary
Apply univariate statistical tests (parametric ANOVA or non-parametric alternatives) to LC-MS/MS metabolomic feature tables to identify metabolites with statistically significant abundance differences across sample classes, followed by multiple-testing correction and feature selection based on q-value thresholds.
When to use
You have a normalized abundance matrix from LC-MS/MS profiling with sample class assignments (e.g., phenotypic groups, disease states, treatment conditions) and need to filter metabolic features for downstream pathway analysis or biological validation. Use this skill when you have ≥3 sample classes to compare and wish to reduce the feature set to those with q-values (FDR-corrected p-values) below a predefined significance cutoff (e.g., q < 0.05).
When NOT to use
- Input is already a filtered list of known biomarkers or curated metabolites; re-testing may inflate false-positive rates.
- Sample class groups are highly imbalanced (e.g., n=1 vs n=100) without appropriate variance-stabilization; consider post-hoc power analysis.
- Feature abundance data are raw, unnormalized counts; apply normalization and quality control filtering (e.g., mass defect filtering, CV thresholding) before statistical testing.
Inputs
- Normalized LC-MS/MS feature abundance matrix (e.g., Urine_RP_NEG_norm.txt or Urine_RP_POS_norm.txt)
- Sample metadata with class/group assignments (e.g., phenotype labels AA, DD, MM)
Outputs
- Filtered feature table with Feature_ID, metabolite names, ANOVA F-statistics, p-values, q-values, and effect sizes
- List of significant metabolic features passing the q-value threshold
How to apply
Load the normalized feature abundance table (e.g., Urine_RP_NEG_norm.txt) into margheRita as a data object with sample metadata annotated with class assignments (AA, DD, MM). Apply the univariate() function to compute ANOVA F-statistics and p-values across all class levels for each metabolite feature; Benjamini–Hochberg FDR correction is applied automatically during this step to yield q-values. Use select_sign_features() to filter significant features by specifying a q-value threshold (commonly 0.05 or 0.01). Extract the resulting table containing Feature_ID, metabolite names, ANOVA statistics, p-values, q-values, and effect sizes for downstream interpretation. The choice between parametric (ANOVA) and non-parametric tests depends on whether feature abundance distributions meet normality assumptions, which should be assessed prior to testing.
Related tools
- margheRita (R package providing univariate() and select_sign_features() functions for ANOVA testing and q-value-based feature filtering on normalized LC-MS/MS data) — https://github.com/emosca-cnr/margheRita
- R (Programming language in which margheRita is implemented)
- notame (Alternative R workflow package for non-targeted LC-MS metabolic profiling with statistical analysis functions; compatible via margheRita export functions) — https://github.com/hanhineva-lab/notame
- MS-DIAL (Peak picking and feature detection software upstream of margheRita; output processed by statistical testing workflow)
Examples
univariate(data_object, metadata_col='class') |> select_sign_features(q_threshold=0.05)
Evaluation signals
- Q-value distribution shows expected enrichment of small values (< 0.05) in features with biological effect, and q-values increase monotonically with increasing p-values after FDR correction.
- Number of significant features (q < threshold) is interpretable relative to study design and metabolite coverage; extreme counts (e.g., >90% of features or <1%) warrant review of threshold or data quality.
- Effect sizes (e.g., fold-change, eta-squared) for significant features are consistent with magnitude of class-level differences visible in exploratory plots (e.g., PCA, box plots by group).
- Replicate runs of univariate() on the same input data yield identical p-values and q-values.
- Tabulated results include all expected columns (Feature_ID, metabolite name, ANOVA F-statistic, p-value, q-value) with no missing or infinite values for passing features.
Limitations
- ANOVA assumes homogeneity of variance across groups; violations may inflate Type I error rates. Consider Welch's ANOVA or non-parametric tests (Kruskal–Wallis) when variances differ substantially.
- Multiple-testing correction (Benjamini–Hochberg FDR) controls false-discovery rate but not family-wise error; choice of q-value threshold (0.05 vs 0.01) is arbitrary and should be justified a priori or validated by independent cohort.
- Statistical significance does not imply biological significance; effect sizes and fold-changes must be examined in context of biological plausibility and prior knowledge.
- Feature abundance distributions may not meet normality assumptions, especially after preprocessing (e.g., log-transformation, imputation of missing values); visualize distributions and consider rank-based tests if violations are severe.
Evidence
- [other] Apply univariate() function to compute ANOVA F-statistics and p-values across the three class levels for each metabolite feature.: "Apply univariate() function to compute ANOVA F-statistics and p-values across the three class levels for each metabolite feature."
- [other] Use select_sign_features() to filter significant features based on q-value threshold (Benjamini–Hochberg FDR correction applied during univariate or select_sign_features step).: "Use select_sign_features() to filter significant features based on q-value threshold (Benjamini–Hochberg FDR correction applied during univariate or select_sign_features step)."
- [other] Extract and tabulate the results including feature identifiers, ANOVA statistics, p-values, q-values, and effect sizes.: "Extract and tabulate the results including feature identifiers, ANOVA statistics, p-values, q-values, and effect sizes."
- [readme] simplified execution of parametric and non-parametric statistical tests over a large number of features: "simplified execution of parametric and non-parametric statistical tests over a large number of features"
- [other] ANOVA testing across AA, DD, and MM class levels in the Urine dataset identified significant metabolic features, with results reported in a table including Feature_ID, metabolite names, and associated q-values for filtering at specified cutoffs.: "ANOVA testing across AA, DD, and MM class levels in the Urine dataset identified significant metabolic features, with results reported in a table including Feature_ID, metabolite names, and"
1---2name: mass-spectrometry-data-statistical-testing3description: Use when you have a normalized abundance matrix from LC-MS/MS profiling with sample class assignments (e.g., phenotypic groups, disease states, treatment conditions) and need to filter metabolic features for downstream pathway analysis or biological validation.4license: CC-BY-4.05---67# mass-spectrometry-data-statistical-testing89## Summary1011Apply univariate statistical tests (parametric ANOVA or non-parametric alternatives) to LC-MS/MS metabolomic feature tables to identify metabolites with statistically significant abundance differences across sample classes, followed by multiple-testing correction and feature selection based on q-value thresholds.1213## When to use1415You have a normalized abundance matrix from LC-MS/MS profiling with sample class assignments (e.g., phenotypic groups, disease states, treatment conditions) and need to filter metabolic features for downstream pathway analysis or biological validation. Use this skill when you have ≥3 sample classes to compare and wish to reduce the feature set to those with q-values (FDR-corrected p-values) below a predefined significance cutoff (e.g., q < 0.05).1617## When NOT to use1819- Input is already a filtered list of known biomarkers or curated metabolites; re-testing may inflate false-positive rates.20- Sample class groups are highly imbalanced (e.g., n=1 vs n=100) without appropriate variance-stabilization; consider post-hoc power analysis.21- Feature abundance data are raw, unnormalized counts; apply normalization and quality control filtering (e.g., mass defect filtering, CV thresholding) before statistical testing.2223## Inputs2425- Normalized LC-MS/MS feature abundance matrix (e.g., Urine_RP_NEG_norm.txt or Urine_RP_POS_norm.txt)26- Sample metadata with class/group assignments (e.g., phenotype labels AA, DD, MM)2728## Outputs2930- Filtered feature table with Feature_ID, metabolite names, ANOVA F-statistics, p-values, q-values, and effect sizes31- List of significant metabolic features passing the q-value threshold3233## How to apply3435Load the normalized feature abundance table (e.g., Urine_RP_NEG_norm.txt) into margheRita as a data object with sample metadata annotated with class assignments (AA, DD, MM). Apply the univariate() function to compute ANOVA F-statistics and p-values across all class levels for each metabolite feature; Benjamini–Hochberg FDR correction is applied automatically during this step to yield q-values. Use select_sign_features() to filter significant features by specifying a q-value threshold (commonly 0.05 or 0.01). Extract the resulting table containing Feature_ID, metabolite names, ANOVA statistics, p-values, q-values, and effect sizes for downstream interpretation. The choice between parametric (ANOVA) and non-parametric tests depends on whether feature abundance distributions meet normality assumptions, which should be assessed prior to testing.3637## Related tools3839- **margheRita** (R package providing univariate() and select_sign_features() functions for ANOVA testing and q-value-based feature filtering on normalized LC-MS/MS data) — https://github.com/emosca-cnr/margheRita40- **R** (Programming language in which margheRita is implemented)41- **notame** (Alternative R workflow package for non-targeted LC-MS metabolic profiling with statistical analysis functions; compatible via margheRita export functions) — https://github.com/hanhineva-lab/notame42- **MS-DIAL** (Peak picking and feature detection software upstream of margheRita; output processed by statistical testing workflow)4344## Examples4546```47univariate(data_object, metadata_col='class') |> select_sign_features(q_threshold=0.05)48```4950## Evaluation signals5152- Q-value distribution shows expected enrichment of small values (< 0.05) in features with biological effect, and q-values increase monotonically with increasing p-values after FDR correction.53- Number of significant features (q < threshold) is interpretable relative to study design and metabolite coverage; extreme counts (e.g., >90% of features or <1%) warrant review of threshold or data quality.54- Effect sizes (e.g., fold-change, eta-squared) for significant features are consistent with magnitude of class-level differences visible in exploratory plots (e.g., PCA, box plots by group).55- Replicate runs of univariate() on the same input data yield identical p-values and q-values.56- Tabulated results include all expected columns (Feature_ID, metabolite name, ANOVA F-statistic, p-value, q-value) with no missing or infinite values for passing features.5758## Limitations5960- ANOVA assumes homogeneity of variance across groups; violations may inflate Type I error rates. Consider Welch's ANOVA or non-parametric tests (Kruskal–Wallis) when variances differ substantially.61- Multiple-testing correction (Benjamini–Hochberg FDR) controls false-discovery rate but not family-wise error; choice of q-value threshold (0.05 vs 0.01) is arbitrary and should be justified a priori or validated by independent cohort.62- Statistical significance does not imply biological significance; effect sizes and fold-changes must be examined in context of biological plausibility and prior knowledge.63- Feature abundance distributions may not meet normality assumptions, especially after preprocessing (e.g., log-transformation, imputation of missing values); visualize distributions and consider rank-based tests if violations are severe.6465## Evidence6667- [other] Apply univariate() function to compute ANOVA F-statistics and p-values across the three class levels for each metabolite feature.: "Apply univariate() function to compute ANOVA F-statistics and p-values across the three class levels for each metabolite feature."68- [other] Use select_sign_features() to filter significant features based on q-value threshold (Benjamini–Hochberg FDR correction applied during univariate or select_sign_features step).: "Use select_sign_features() to filter significant features based on q-value threshold (Benjamini–Hochberg FDR correction applied during univariate or select_sign_features step)."69- [other] Extract and tabulate the results including feature identifiers, ANOVA statistics, p-values, q-values, and effect sizes.: "Extract and tabulate the results including feature identifiers, ANOVA statistics, p-values, q-values, and effect sizes."70- [readme] simplified execution of parametric and non-parametric statistical tests over a large number of features: "simplified execution of parametric and non-parametric statistical tests over a large number of features"71- [other] ANOVA testing across AA, DD, and MM class levels in the Urine dataset identified significant metabolic features, with results reported in a table including Feature_ID, metabolite names, and associated q-values for filtering at specified cutoffs.: "ANOVA testing across AA, DD, and MM class levels in the Urine dataset identified significant metabolic features, with results reported in a table including Feature_ID, metabolite names, and"
HolobiomicsLab