Easy access to benchmark datasets.
To add a dataset, just create a section with a description and links to download it below. How easy can you make it to get started?
A discussion of the problems for which benchmark datasets would allow for experimentation.
- cell type annotation and reannotation at various levels of ontological depth
- building and validating cell type classifiers
- manifold alignment and batch-effect-aware analyses
- assessing the variability in gene expression of cell types present in many organs
- measuring sex differences in gene expression
- measuring the variability in biological claims (like which genes are differentially expressed between populations) to be expected between different studies of the same cell types
Tabula Muris contains about 100,000 cells from 20 organs and tissues in mouse. The study is sex-balanced, with four male and four female mice. The organs included are skin, fat, mammary gland, heart, bladder, brain, thymus, spleen, kidney, limb muscle, tongue, marrow, trachea, pancreas, lung, large intestine, and liver. Many of these organs were processed using two methods: SMART-seq2 on FACS-sorted cells and microfluidic droplets from 10X Genomics.
Below are instructions for getting four files: metadata (including annotations) and count data for each dataset.
Version-controlled metadata are available on github.
You can download complete count files as sparse matrices in .rds format for easy loading into R. Unzip TabulaMuris.zip. Load:
tm.droplet.matrix = readRDS(here("data", "TM_droplet_mat.rds"))
tm.droplet.metadata = read_csv(here("data", "TM_droplet_metadata.csv"))You can download complete count files as sparse matrices in AnnData-formatted h5ad files for use in Python here. You can load them using the Scanpy library:
import pandas
import scanpy
tm_facs_metadata = pd.read_csv('data/TM_facs_metadata.csv')
tm_facs_data = scanpy.anndata.read_h5ad('data/TM_facs_mat.h5ad')The original data release is on FigShare.