Dockerized MAST (Model-based Analysis of Single-cell Transcriptomics) for computing genes that are differentially expressed in each cluster compared to the rest of the data.

prepare-MAST.ipynb takes a h5ad file and generates:

df.0.RDS
df.0.rest.RDS
df.0.tgz
.
.
.
df.n.RDS
df.n.rest.RDS
df.n.tgz

where 0..n is the cluster number.

df.n.tgz is a tarball gzipped of two files: df.n.RDS and df.n.rest.RDS.

The reason is to analyze each cluster in parallel way (rather than passing the entire h5ad or RDS to each worker).

pairwise-diffexp.R takes two arguments:

• cluster number
• output CSV filename

For example, if you specify cluster number 5, it will load df.1.RDS and df.1.rest.RDS and compute the genes that are differentially expressed in cluster 5 compared to the rest of the data.

./run-notebook.sh

./build.sh

Either you can use the docker push command or run push.sh (requires SCING):

./push.sh