Background 

Understanding cell-to-cell communication mediated by ligand-receptor complexes is crucial for comprehending biological processes like development, differentiation, and inflammation. Investigating the crosstalk between different cell types provides insights into how physiological processes are facilitated. The scalability of single-cell genomics technologies has allowed for parallel analysis of multiple cells from the same tissue, providing a unique opportunity to study cell communication. However, existing tools for estimating ligand-receptor interactions faced two major issues: (i) the majority of ligands and receptors stored were not curated, leading to a lot of false positives (i.e. interactions stored in a database that do not exist in the reality); (ii) the existing methods did not account for the multimer nature of the majority of the ligand-receptor interactions, meaning each receptor (or ligand) consist by more than one subunit. The latter is important to make accurate predictions as it is the combination of subunits that gives specificity to the ligand/receptor interactions. To address these challenges, Dr Roser Vento-Tormo, group leader in the VenTo Lab, at Wellcome Sanger Institute, and collaborators developed CellphoneDB.

Function

CellphoneDB is a suite to study cell-to-cell communication from single-cell transcriptomics data. Identifying ligand–receptor interactions from single-cell RNA sequencing (scRNA-seq) requires both the annotation of complex ligand–receptor relationships from the literature (i.e. database) and a statistical method that integrates the resource with scRNA-seq data and selects relevant interactions from the dataset (ie. tool). CellphoneDB is composed of two units, a database and a tool. CellphoneDB database is a publicly available repository of curated receptors, ligands and their interactions. The database can be used to search for a particular ligand/receptor or in combination with the tool to interrogate your own single-cell transcriptomics data.

Development process

The development process was a highly collaborative and multidisciplinary work, that started in the Teichmann lab (additional information can be found here). “The computational approach was discussed multiple times in the lab meetings where multiple members from the Teichmann lab contributed to choose the best statistical approach used to infer ligand-receptor interactions.”, said Dr Vento-Tormo. The team also worked closely with experts in the field to ensure that receptor and ligand interactions stored in their database were correct. That included Gerry Graham, Professor at the University of Glasgow (additional information can be found here), an expert on cell-cell communication through chemokines. In addition, the group established collaborations with software developers that enabled to build the structure of the database, make the code scalable and have a webpage to query the ligand/receptor interactions present in the database. 

One of the most challenging part in the development process was the scalability, explains Dr Vento-Tormo. “Since 2016, when we started the method, the amount of single cell generated has increased significantly and we had to account for this. Also, it was surprising how little we know about protein complexes (i.e. multimers) which are essential to understand receptors.”

Target user

In the first instance researchers. Also, anyone who wants to make their own synthetic sitting pathways, try some new combinations or expressing these new cell types, or using it for some new technology. CellphoneDB has a lot of translational potential and non-academic researchers are becoming more interested in it because the receptors are surface proteins that can be targeted with biologics.

Comparison to other technologies

The first differential of CellphoneDB database is that it only includes curated data, meaning that only interactions linked to a manuscript are included. Secondly, the database considers that the majority of receptors are multimers: in other words, that multiple subunits are required in order for the receptor to be functional. Thirdly, in later versions CellphoneDB also includes information about: a) non-protein ligands (e.g. steroid hormones), by taking into account the last enzyme involved in the production of the non-protein ligand; b) transcription factor and receptors that are part of the same signalling pathway. The link with downstream transcription factors, could be used as a sensor of the ligand-receptor interaction.  Finally, the database and tool are continuously updated and made compatible with other tools to analyse single-cell genomics data.