Try the Demo
Load one of the example datasets below to explore the app before uploading your own data.
Ionization Modes
Positive and negative mode acquisitions are separate experiments: intensities are not comparable between them, so each is normalized, filtered and batch corrected on its own. They are brought together only for differential analysis, where a single FDR correction is applied across both.
Mode 1
Upload Feature Data — Mode 1
Preview:
Upload Sample Metadata — Mode 1
Preview:
Mode 2
Upload Feature Data — Mode 2
Preview:
Upload Sample Metadata — Mode 2
Preview:
Configure Feature Data — Mode 1
Configure Sample Data — Mode 1
Step 1: Preview Sample Matching — Mode 1
Step 2: Preprocessing — Mode 1
Filtering and imputation:
Sample normalization:
Log2 transformation will always be applied after sample normalization.
Step 3: PCA Before Batch Correction — Mode 1
Configure Feature Data — Mode 2
Configure Sample Data — Mode 2
Step 1: Preview Sample Matching — Mode 2
Step 2: Preprocessing — Mode 2
Filtering and imputation:
Sample normalization:
Log2 transformation will always be applied after sample normalization.
Step 3: PCA Before Batch Correction — Mode 2
About this step
Batch correction is optional. If the PCA on the previous tab shows no separation by batch, there is nothing to remove — skip ComBat and apply scaling only. Differential analysis can still account for batch by including it as a covariate, which is the safer choice when a batch effect is small or uncertain.
Scaling is applied here either way, because the scaled matrices are what PCA, clustering and heatmaps use.
Batch Correction & Scaling — Mode 1
Run preprocessing for this mode first.
Scaling is applied to the normalized data, and to the batch-corrected data as well when ComBat is run. Scaling affects PCA, clustering and heatmaps; differential analysis is always modeled on the unscaled log2 data.
PCA after batch correction and scaling:
Batch Correction & Scaling — Mode 2
Run preprocessing for this mode first.
Scaling is applied to the normalized data, and to the batch-corrected data as well when ComBat is run. Scaling affects PCA, clustering and heatmaps; differential analysis is always modeled on the unscaled log2 data.
PCA after batch correction and scaling:
Set Up Comparison
Run preprocessing on the Preprocess Data tab first. Differential analysis needs the log2-normalized matrix.
What to compare:
Groups:
Log2 fold change is A minus B: positive values are higher in group A.
Groups to include:
The F-test asks whether a feature differs across any of the selected groups. Unselected groups are left out of the model entirely, so they do not affect the variance estimate.
Data and thresholds:
Results
Saved Comparisons
Every comparison you run is saved and included in the Excel package and HTML report. Re-running the same comparison replaces its saved copy.
About Metabo Tools
Metabo Tools is a web application for metabolomics data preprocessing, batch correction, and differential analysis.
This tool is currently under active development and testing. Please use results with appropriate caution and report any issues.
Source Code
The source code is available on GitHub:
https://github.com/UFHCC-BCBSR/app-metab-tools
Contact
For questions, feedback, or to report issues, contact: hkates@ufl.edu
Developed in partnership with SECIM
This tool was developed in partnership with the University of Florida Southeast Center for Integrated Metabolomics (SECIM) .
University of Florida Health Cancer Center — Biostatistics, Computational Biology, and Bioinformatics Shared Resource (BCBSR)
Download Results
Download your complete results package. The Excel file contains every processed data version for each ionization mode, a README sheet, and one sheet per differential analysis comparison. The HTML report contains a methods paragraph, processing log, file guide and PCA plots for each mode, plus a full section for every comparison you ran.
Download Excel Data Package
Download HTML Report
Complete preprocessing and the scaling step for at least one mode to enable downloads.