Cachexia Atlas
Published cachexia data, one gene at a time
A public Shiny app for browsing published cancer-cachexia datasets in mouse and human skeletal muscle. It is built to make cross-dataset comparison possible without implying the datasets are comparable.
What it is
The atlas collects published cancer-cachexia datasets across mouse and human skeletal muscle and puts them behind one interface, at both the transcript and the protein level. Assays span bulk RNA-seq, microarray, single-cell and single-nucleus RNA-seq, and mass-spectrometry proteomics.
What you can do
Search a gene
See it across cell types, sample groups, or protein groups, depending on what the dataset measured.
Open one dataset
Each dataset has a full explorer, with differential analysis and pathway enrichment.
Compare across datasets
One gene across every dataset on one screen.
How it reports uncertainty
Nothing here was measured on the same material twice. Every cross-dataset comparison is hypothesis-generating rather than a multi-omic measurement, and the app says so instead of hiding it.
The gene-across-datasets view defaults to a sign-and-significance strip, not a common effect-size axis. Putting non-comparable datasets on one axis would imply a comparability that does not exist.
"Not measured" and "not detected" are different states, and they are drawn differently rather than collapsing into one blank cell.
- Not measured
- outside the assay
- Not detected
- assayed, no signal
Where the atlas recomputes a statistic instead of using the published one, it is labeled as recomputed.
Agreement in direction across datasets is not replication. Two studies can point the same way for reasons that have nothing to do with the same biology.
Stack
- R Shiny + bslib
- one module per section
- YAML dataset manifests
- validated against a schema
- on-disk HDF5 for single-cell
- Docker on Railway