Data Preparation
Load a feature quantification table (features in rows, samples in columns)
and a metadata table (one row per sample, with a filename column).
Then clean up, impute, normalize and scale — everything runs locally, your files never upload.
1 · Load your two tables
.mzML.CSV / TSV / TXT.
filename column plus attribute columns.CSV / TSV / TXT.
Principal Component Analysis
Unsupervised projection of samples onto the axes of greatest variance. Points are coloured by your chosen attribute.
Scores plot
Explained variance (scree)
PCoA & PERMANOVA
Principal coordinates analysis on a sample-to-sample distance matrix, with a PERMANOVA permutation test for whether groups differ.
Distance metric
PCoA ordination
PERMANOVA result
Hierarchical Clustering & Heatmap
The most variable features across samples, z-scored per feature, with samples and features reordered by average-linkage hierarchical clustering.
Options
Heatmap
Univariate Statistics
Per-feature group comparison. With two groups a Welch t-test is used; with more, one-way ANOVA — or Kruskal–Wallis for a non-parametric test. p-values are Benjamini–Hochberg corrected.
Test
Results — click a feature for its box plot
A browser-only companion to the FBMN-STATS Guide (Functional Metabolomics Lab). Methods are hand-implemented for transparency and offline use; for publication-grade pipelines with QIIME2/R equivalents, batch correction and molecular-network integration, use the original notebooks and web app. Nothing here is uploaded — all computation is local.