Tutorials

These are the six tutorials of the EcoTyper README, run with the Python port on the example data bundled with the R repository. Each one opens with the short scverse-API version (et.recovery / et.discovery), then gives the pandas route, the file-based route that writes the same files as R, and what to look at in the results.

Before you start: install the package and clone the R repository for its example data and prebuilt models (Installation › Example data). All paths below are relative to the folder that contains the ecotyper/ clone.

#	Tutorial	Data	Model	Python time*
1	Recovery in bulk data	TCGA lung (250 samples), DLBCL (75 samples)	Carcinoma, Lymphoma	< 1 min
2	Recovery in scRNA-seq data	colorectal cancer (13,780 cells), lymphoma (1,000 cells)	Carcinoma, Lymphoma	~3 min
3	Recovery in Visium spatial data	breast cancer Visium (3,813 spots)	Carcinoma	~6 min
4	Discovery in bulk data	CIBERSORTx-purified lung profiles	new model	~11 min
5	Discovery in scRNA-seq data	colorectal cancer scRNA-seq	new model	~3.4 h
6	Discovery from presorted profiles	presorted per-cell-type profiles	new model	~11 min

* Wall time on a 112-core x86 server with 32 parallel jobs. See Benchmarks. With fewer jobs, discovery scales roughly linearly with the number of (cell type, rank) NMF runs per job.