diff --git a/assays/ANOVAForDifferentialEnrichmentOfMetabolites/README.md b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/assays/ANOVAForDifferentialEnrichmentOfMetabolites/dataset/.gitkeep b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/dataset/.gitkeep
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/assays/ANOVAForDifferentialEnrichmentOfMetabolites/isa.assay.xlsx b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/isa.assay.xlsx
new file mode 100644
index 0000000000000000000000000000000000000000..4f77dadc2ab5adcbbe4f3acac490e4d67fcafd5a
Binary files /dev/null and b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/isa.assay.xlsx differ
diff --git a/assays/ANOVAForDifferentialEnrichmentOfMetabolites/protocols/.gitkeep b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/protocols/.gitkeep
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/assays/ANOVAForDifferentialEnrichmentOfMetabolites/protocols/ANOVAForDifferentialEnrichmentOfMetabolitesProtocol.md b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/protocols/ANOVAForDifferentialEnrichmentOfMetabolitesProtocol.md
new file mode 100644
index 0000000000000000000000000000000000000000..941712ccb91229a3a43a4fba1216fe1b009d7ff4
--- /dev/null
+++ b/assays/ANOVAForDifferentialEnrichmentOfMetabolites/protocols/ANOVAForDifferentialEnrichmentOfMetabolitesProtocol.md
@@ -0,0 +1,3 @@
+## ANOVA for differential enrichment of metabolites
+
+We applied the python package *statsmodels* (Seabold, S. and Perktold, J., 2010) to perform two-way analysis of variance (ANOVA) (Girden, E., 1992). For a particular metabolite, we collected its mass spectrometry relative response in the WT or in sweet knockouts (line: WT vs. SWEET), and at different segments of the root (region: 2cm vs. 4cm vs. 6cm vs. whole). We fit a model that decomposes the relative response of the metabolite as a combination of line-specific and region-specific factors; this model is described by the following string in the code: *‘relativeResponse ∼ Line + Region'* and calculates a *P*-value for *Line* and one for *Region*. Using the *P*-value cutoff of 0.05 for *Line* (and *Region*), we determined the relative response of the metabolite to change significantly with *Line* (*Region*) if *P* <0.05.
\ No newline at end of file