First compiled: May 22, 2017
This uses the publicly available 10X 1.3 mio brain cell data set.
The 1000 most variable genes.
Currently, this runs about four hours. Note the continuous structures.
Currently, this runs about 10 hours. It will be much faster, soon. Note the clear structures, directions of probability flow.
This has been computed using Scanpy's command-line interface, where the function used to initialize the run was
def one_m():
filename = "path_to_file"
genome = "mm10"
adata = sc.read_10x_h5(filename, genome)
sc.pp.recipe_zheng17(adata)
sc.pp.pca(adata, n_comps=50, zero_center=False)
return adata
Simply use this in a file runs.py in your current working directory and type
scanpy one_m tsne
scanpy one_m diffmap
on the command line. This directly calls the API functions with the same name. See scanpy --help and scanpy tsne --help for information on how to change parameters.