metabolomics-intensity-normalization
License: noncommercial — confirm your use is a permitted (noncommercial) purpose before applying; commercial use requires a separate license (see metadata.tool_license).
Summary
Apply robust normalization methods (tGAM, rGAM, rLOESS, QC-RLSC, QC-RSC) to correct for signal drift and batch effects in untargeted metabolomics intensity matrices. This skill removes systematic variation across analytical runs while preserving biological signal, essential for cross-batch and multi-scale metabolomics studies.
When to use
Your input is a raw metabolomics intensity matrix (compounds × samples) with known batch assignment and QC sample labels, and you observe signal drift across the analytical sequence or batch-to-batch variation. Apply this skill before statistical testing or multivariate analysis to enable fair comparison of biological samples.
When NOT to use
- Input is already normalized or has had batch effects removed by other methods.
- You have no QC samples or batch labels in your experimental design.
- Your data are already feature-level (e.g., peak areas with retention time alignment complete); normalization requires raw intensity values prior to feature extraction.
Inputs
- Unnormalized metabolomics intensity matrix (numerical matrix: rows = compounds, columns = samples)
- Batch assignment vector (character or factor, one entry per sample)
- QC sample type vector (character or factor, entries: 'QC' or other sample type labels)
Outputs
- Normalized metabolomics intensity matrix (same dimensions as input)
- Diagnostic plots: intensity vs. run order pre- and post-normalization (per compound)
- PCA score plots before and after normalization (labeled by batch)
How to apply
Load the raw intensity matrix (rows = compounds, columns = samples), batch vector, and QC sample type labels into R. Install Metanorm from GitHub and call the metanorm() function with your chosen model (tGAM recommended for superior robustness; rGAM or rLOESS for faster computation). Set QCcheck=TRUE to verify QC samples are representative of the batch. Optionally set QConly=TRUE if using legacy methods (QC-RLSC, QC-RSC) that require QC samples only; for tGAM, rGAM, and rLOESS, include both QC and biological samples for normalization. Generate diagnostic plots (intensity vs. run order, pre/post PCA) to visually confirm drift correction and assess whether batch effects have diminished. Export the normalized intensity matrix for downstream analysis.
Related tools
Examples
library(metanorm); load(system.file('extdata', 'example.RData', package = 'metanorm')); normdat <- metanorm(rawdata[1:5,], model = 'tGAM', type = metanorm.qc, QCcheck = TRUE, batch = batch, plotdir = '~/metanormExample/')
Evaluation signals
- PCA score plots show reduced or eliminated clustering by batch after normalization (visual confirmation of drift correction).
- Individual compound intensity vs. run order plots exhibit flattened trend lines post-normalization, with stable mean intensity across the analytical sequence.
- QC sample intensities remain consistent pre- and post-normalization (QC samples should not be artificially altered by the normalization).
- Biological sample relative abundances are preserved: fold-change ratios between sample groups remain stable or improve in signal-to-noise ratio.
- Metanorm QCcheck diagnostic output confirms QC samples are representative (no flagged discrepancies between QC and biological sample distributions).
Limitations
- tGAM offers superior robustness but may be slower than rGAM or rLOESS; runtimes scale with number of compounds and samples.
- Normalization assumes systematic drift is smooth and monotonic or quasi-periodic; sharp instrumental failures or abrupt baseline shifts may not be fully corrected.
- Results depend on the representativeness of QC samples; if QC samples do not span the chemical space of biological samples, normalization may be suboptimal.
- The method is designed for untargeted metabolomics; applicability to targeted assays or non-LC-MS platforms has not been evaluated in the cited work.
Evidence
- [readme] Metanorm supports robust metabolomics data normalization across scales and experimental designs. The R package implements three (new) robust normalization methods (tGAM, rGAM and rLOESS), alongside formerly proposed ones (QC-RLSC, QC-RSC).: "Metanorm supports robust metabolomics data normalization across scales and experimental designs. The R package implements three (new) robust normalization methods (tGAM, rGAM and rLOESS), alongside"
- [readme] tGAM is recommended due to its superior robustness, but rGAM and rLOESS are faster.: "tGAM is recommended due to its superior robustness, but rGAM and rLOESS are faster."
- [readme] We further recommend using both QC as well as biological samples for normalization, and to have metanorm check for discrepancies between QC and biological samples.: "We further recommend using both QC as well as biological samples for normalization, and to have metanorm check for discrepancies between QC and biological samples."
- [readme] Normalizing the data is achieved by calling the metanorm function. normalize the first 5 compounds in the example dataset using the (default) tGAM method normdat <- metanorm(rawdata[1:5,], model = 'tGAM', type = metanorm.qc, QCcheck = TRUE, batch = batch, plotdir = '~/Documents/metanormExample/'): "normalize the first 5 compounds in the example dataset using the (default) tGAM method normdat <- metanorm(rawdata[1:5,], model = 'tGAM', type = metanorm.qc, QCcheck = TRUE, batch = batch, plotdir ="
- [readme] Individual compound pre- vs. post-normalization intensity vs. order plots can be retrieved from the plotdir directory. These allow finegrained assessment of normalization performance.: "Individual compound pre- vs. post-normalization intensity vs. order plots can be retrieved from the plotdir directory. These allow finegrained assessment of normalization performance."
1---2name: metabolomics-intensity-normalization-23description: Use when your input is a raw metabolomics intensity matrix (compounds × samples) with known batch assignment and QC sample labels, and you observe signal drift across the analytical sequence or batch-to-batch variation.4license: CC-BY-4.05---67# metabolomics-intensity-normalization89> **License: noncommercial** — confirm your use is a permitted (noncommercial) purpose before applying; commercial use requires a separate license (see `metadata.tool_license`). <!-- asb-license-banner -->10## Summary1112Apply robust normalization methods (tGAM, rGAM, rLOESS, QC-RLSC, QC-RSC) to correct for signal drift and batch effects in untargeted metabolomics intensity matrices. This skill removes systematic variation across analytical runs while preserving biological signal, essential for cross-batch and multi-scale metabolomics studies.1314## When to use1516Your input is a raw metabolomics intensity matrix (compounds × samples) with known batch assignment and QC sample labels, and you observe signal drift across the analytical sequence or batch-to-batch variation. Apply this skill before statistical testing or multivariate analysis to enable fair comparison of biological samples.1718## When NOT to use1920- Input is already normalized or has had batch effects removed by other methods.21- You have no QC samples or batch labels in your experimental design.22- Your data are already feature-level (e.g., peak areas with retention time alignment complete); normalization requires raw intensity values prior to feature extraction.2324## Inputs2526- Unnormalized metabolomics intensity matrix (numerical matrix: rows = compounds, columns = samples)27- Batch assignment vector (character or factor, one entry per sample)28- QC sample type vector (character or factor, entries: 'QC' or other sample type labels)2930## Outputs3132- Normalized metabolomics intensity matrix (same dimensions as input)33- Diagnostic plots: intensity vs. run order pre- and post-normalization (per compound)34- PCA score plots before and after normalization (labeled by batch)3536## How to apply3738Load the raw intensity matrix (rows = compounds, columns = samples), batch vector, and QC sample type labels into R. Install Metanorm from GitHub and call the metanorm() function with your chosen model (tGAM recommended for superior robustness; rGAM or rLOESS for faster computation). Set QCcheck=TRUE to verify QC samples are representative of the batch. Optionally set QConly=TRUE if using legacy methods (QC-RLSC, QC-RSC) that require QC samples only; for tGAM, rGAM, and rLOESS, include both QC and biological samples for normalization. Generate diagnostic plots (intensity vs. run order, pre/post PCA) to visually confirm drift correction and assess whether batch effects have diminished. Export the normalized intensity matrix for downstream analysis.3940## Related tools4142- **Metanorm** (R package implementing five robust normalization methods (tGAM, rGAM, rLOESS, QC-RLSC, QC-RSC) and diagnostic plotting functions for metabolomics intensity data) — https://github.com/UGent-LIMET/Metanorm43- **R** (Runtime environment for executing Metanorm and data manipulation workflows (≥4.4.0 required)) — https://cloud.r-project.org/index.html4445## Examples4647```48library(metanorm); load(system.file('extdata', 'example.RData', package = 'metanorm')); normdat <- metanorm(rawdata[1:5,], model = 'tGAM', type = metanorm.qc, QCcheck = TRUE, batch = batch, plotdir = '~/metanormExample/')49```5051## Evaluation signals5253- PCA score plots show reduced or eliminated clustering by batch after normalization (visual confirmation of drift correction).54- Individual compound intensity vs. run order plots exhibit flattened trend lines post-normalization, with stable mean intensity across the analytical sequence.55- QC sample intensities remain consistent pre- and post-normalization (QC samples should not be artificially altered by the normalization).56- Biological sample relative abundances are preserved: fold-change ratios between sample groups remain stable or improve in signal-to-noise ratio.57- Metanorm QCcheck diagnostic output confirms QC samples are representative (no flagged discrepancies between QC and biological sample distributions).5859## Limitations6061- tGAM offers superior robustness but may be slower than rGAM or rLOESS; runtimes scale with number of compounds and samples.62- Normalization assumes systematic drift is smooth and monotonic or quasi-periodic; sharp instrumental failures or abrupt baseline shifts may not be fully corrected.63- Results depend on the representativeness of QC samples; if QC samples do not span the chemical space of biological samples, normalization may be suboptimal.64- The method is designed for untargeted metabolomics; applicability to targeted assays or non-LC-MS platforms has not been evaluated in the cited work.6566## Evidence6768- [readme] Metanorm supports robust metabolomics data normalization across scales and experimental designs. The R package implements three (new) robust normalization methods (tGAM, rGAM and rLOESS), alongside formerly proposed ones (QC-RLSC, QC-RSC).: "Metanorm supports robust metabolomics data normalization across scales and experimental designs. The R package implements three (new) robust normalization methods (tGAM, rGAM and rLOESS), alongside"69- [readme] tGAM is recommended due to its superior robustness, but rGAM and rLOESS are faster.: "tGAM is recommended due to its superior robustness, but rGAM and rLOESS are faster."70- [readme] We further recommend using both QC as well as biological samples for normalization, and to have metanorm check for discrepancies between QC and biological samples.: "We further recommend using both QC as well as biological samples for normalization, and to have metanorm check for discrepancies between QC and biological samples."71- [readme] Normalizing the data is achieved by calling the metanorm function. normalize the first 5 compounds in the example dataset using the (default) tGAM method normdat <- metanorm(rawdata[1:5,], model = 'tGAM', type = metanorm.qc, QCcheck = TRUE, batch = batch, plotdir = '~/Documents/metanormExample/'): "normalize the first 5 compounds in the example dataset using the (default) tGAM method normdat <- metanorm(rawdata[1:5,], model = 'tGAM', type = metanorm.qc, QCcheck = TRUE, batch = batch, plotdir ="72- [readme] Individual compound pre- vs. post-normalization intensity vs. order plots can be retrieved from the plotdir directory. These allow finegrained assessment of normalization performance.: "Individual compound pre- vs. post-normalization intensity vs. order plots can be retrieved from the plotdir directory. These allow finegrained assessment of normalization performance."
HolobiomicsLab 