RNAseq Bench Webcreated by Paul H. Kim, Ph.D.
runs entirely in your browser · nothing is uploaded
Single-cell QC, clustering and marker genes, computed in this tab. Your files are read by
the page and never sent anywhere — there is no server to send them to.
What this web version can and cannot do. It runs the steps that port to the browser, using the same
libraries the desktop app uses underneath scanpy — scikit-learn for PCA and neighbours, igraph for Leiden,
a rank-sum test for markers. It does not do differential expression between conditions, batch
integration, automated annotation or gene sets: DESeq2, Harmony, scVI, CellTypist and decoupler cannot run
in WebAssembly, and a hand-rolled substitute presented as the real thing would be worse than not having it.
For those, use the desktop app. Practical up to roughly 20–30 thousand cells.
Starting Python
Downloading the scientific stack (~30 MB, cached afterwards)…
1 · Your data
A 10x folder (matrix.mtx + features + barcodes), an
.h5ad, or a counts table (genes as rows). .gz is fine.
Drop files, or click to choose
everything stays on this machine
2 · Quality control
Genes per cell
Counts per cell
Mitochondrial %
Ribosomal %
What the raw data suggests
Most abundant genes
Gene
Mean counts
% cells
Working
…
3 · Result
How this dataset looks
Marker genes
Rank-sum of each cluster against every other cell. A ranking of what marks each cluster —
the p-values behind it are optimistic, because the clusters were defined from these same values.
Cluster
Gene
log2FC
% in
% out
z
RNAseq Bench Web · created by Paul H. Kim, Ph.D. · the desktop app does the statistics this one cannot