Ligand-receptor cellular networks from bulk and single-cell proteomics data

The dialog of cells in a tissue through the secretion of ligands and sensing by receptors plays an essential role in development, homeostasis, and diseases (Ramilowski et al., 2015). The advent of single-cell omics has led to remarkable progresses in the analysis of the cell composition and ligand-receptor networks within a tissue. Nevertheless, single-cell data on patient cohorts remain limited compared with bulk datasets, especially in proteomics. To bring the inference of ligand-receptor interactions (LRIs) to regular expression (bulk) proteomics from tissues, we have introduced a sophisticated R library: BulkSignalR (Villemin et al., 2023). BulkSignalR enables reliable LRI inference through a statistical model that integrates observed correlations between the ligand and the receptor, and between the receptor and downstream pathways over a global, reference molecular network (Fig. 1A). Many analytical and graphical functions are available to represent and further manipulate the obtained cellular networks (Fig. 1BC).

Very recently, we extended BulkSignalR statistical model to also analyze single-cell data, in particular single-cell proteomics. To the best of our knowledge, this is the first LRI tool adapted to single-cell proteomics. Since this new extension of BulkSignalR can also deal with single-cell transcriptomics, we will show results illustrating the complementarity of the two modalities when available for the same samples. In Specht et al. (2021) data, a significant overlap is observed, but each modality uncovers otherwise missed and relevant cellular interactions in comparable proportions, i.e., there is no obvious superiority but clear complementarity (Fig. 1D). This phenomenon certainly relates to molecule abundances but also proteins with longer half-life that do not necessitate constant gene transcription.

Figure 1. (A) Tabular output with triples (ligand, receptor, downstream pathway) along with statistical significance. (B) Classical chord diagram representation of LRIs. (C) Since we model receptor downstream pathways, we can extract networks relating a LRI to its regulatory targets, i.e., to link cellular networks with intracellular networks. (D) Example application of the single-cell extension to Specht et al. data illustrating the complementarity of single-cell transcriptomics and proteomics on such an inter-/intracellular network for differentiated U937 cells.