# Receptor-ligand permutation test¶

This vignette demonstrates the usage of the permutation test as described in CellPhoneDB with the interations extracted from OmniPath database developed by Saezlab.

The downside is that in the original implementation, apart from being inefficient, the CellPhoneDB database has only been manually curated for human interactions. To overcome this issue, we make use of the OmniPath database (containing CellPhoneDB as one of its many sources) which also focuses on literature curated rodent signalling pathways.

## Usage¶

Normalize and create .raw.

## Run the CellPhoneDB's permutation test¶

### Use only CellPhoneDB as a resource¶

For mouse data, CellPhoneDB uses the ortholog genes, downloaded from biomart. They convert the mouse genes into their human orthologs and use that as an input (latest source).

The tori mark significant p-values (alpha=0.001 by default). molecule1 belongs to the source cluster (top) whereas moleule2 to the target clusters.

### Use all available resources from OmniPath¶

Using OmniPath as an interation source yields approx. ~7x more interactions than from CellPhoneDB for the selected human data and ~11x interactions for the seleted mouse data (internally in squidpy.gr.ligrec, we map the mouse gene symbols to human simply by uppercasing).