Your enrichment table, checked before you believe it

Paste the results table your enrichment tool just wrote. Your browser checks what can be checked: the p-values against each row's own counts, the adjusted values against what a correction can produce, the set sizes, the background and the gene IDs. It also groups the redundant terms into themes by shared genes. All free, before you sign in. Then the desk reads the themes in plain biology and writes your results and methods paragraphs, and every number it writes is checked against your table.

Each example comes with a saved model reading, one per verdict (clear, qualified, rerun), so you can see the whole page without signing in or spending a credit.

Your results table (paste or drop it exactly as the tool wrote it, header row included) nothing pasted yet

Drag a .tsv, .csv or .txt in, or Read in your browser. Nothing uploads until you run.

Organism

Cutoff adjusted p

Background universe (optional: what you tested against)

The comparison (which cells or samples, which contrast, and which list or ranking went in)

The gene list you tested (optional: one per line, or GENE score for a ranked list) none

What you want to know (optional)

Table facts (free, in your browser, no model)

What the browser found

Themes (significant terms grouped by shared genes)

Theme	Lead term	Direction	Best adj. p	Terms	Genes

Significant terms

Row	Term	Overlap	p	Adjusted p	NES	Flags

Free: no account, no run.

Write-up draft (from the table only; fill the brackets before you paste)

The reading names each theme in plain biology and answers every flag.

Compare with an earlier run

Pin this table, re-run your tool with the fix, paste the new table: what changed shows here.

Paste a table to price the reading.

significant terms

themes reported

set aside

pitfalls

statements checked

disagreements

Themes

Set aside

Pitfalls

Severity	Code	Finding	Fix	Source

Next steps

Results text

Methods text

Square brackets mark what your table does not state (tool version, library release, seed). Fill them in before you paste.

Reconciliation with your table

Every theme id, representative term, direction, key gene, flag and number the reading wrote was checked against the table facts the browser computed. A disagreement means the reading, not your table, is wrong.

Summary

Raw model reply

The reply did not parse as the structured result, so it is shown as it arrived.

What this does, and what it does not

The table facts are computed, not guessed. When a row carries its overlap, set size, list size and background (clusterProfiler's GeneRatio and BgRatio, DAVID's counts), the browser recomputes the one-sided hypergeometric p-value exactly, from those four numbers alone. Within each library it computes the Benjamini-Hochberg values of the rows you pasted. Adding more tests can only raise those values, so a reported BH value below them cannot be right. Terms are grouped into a theme when they share genes (Jaccard at least 0.5, or at least 80% of the smaller term's genes). The first-ranked term leads the theme.

It does not re-run your enrichment and it has no gene-set database. A table without genes per row can only be grouped by term names. A table without counts can only be checked for internal consistency. The reading explains what a gene set usually means from general knowledge. It does not cite papers or tell you what your experiment proves. Derived from the agent skill @k-dense-ai/pathway-enrichment (K-Dense-AI/scientific-agent-skills, MIT). Not affiliated with Enrichr, g:Profiler, clusterProfiler, gseapy, DAVID or the Broad Institute.

Nothing to hand? Try the : fourteen clusterProfiler rows whose p-values the browser re-derives exactly. Or try the , whose Enrichr table came from a different gene list than the one pasted. Both replay a saved reading for free.