Ligand-receptor interactions were more traditionally analysed using transcriptomics data but now there has been a shift to single-cell proteomics. Jacques Colinge, a Professor at the University of Montpellier, shared his research on the tumour microenvironment and ligand-receptor interactions that take place. Colinge highlighted how to use single-cell transcriptomics to pinpoint different cell types and predict ligand-receptor interactions based on differential expression.
Classical processing in the context of single-cell proteomics data is heavily normalised due to the complicated processing and batch effects. Such statistical approaches end up generating data that resemble Z-scores and do not follow a normal distribution and a loss of information, complicating classical statistical methods.
Colinge briefly talked through a research project analysing monocyte-to-macrophage differentiation. The researchers applied a Z-score-based method to identify autocrine interactions (within the same cell type) as well as interactions between monocytes and macrophages. Colinge mentioned that only 170 ligand-receptor interactions were identified from over 3000 due to the fact that the dataset mostly comprises nuclear and cytoplasmic proteins. Despite the small figure, existing literature supports a number of meaningful interactions, aligning with known differentiation processes.
Switching gears, Colinge focused on bulk data and a recent package for ligand-receptor interactions in a bulk context. Bulk data consists of a mix of information from many cell types, making it difficult to discern clear patterns. So, Colinge developed a new statistical model for bulk omics data that relates the activity of the ligand with the receptor and considers the downstream signal. It ultimately allows the user to measure the activity of a ligand-receptor pair across many samples. The model has been applied to salivary duct carcinoma, discovering interactions relevant to immunotherapy such as new immune checkpoints.
The model also works with spatial transcriptomics data like 10X Visium. In summary, Z-score-based methods can effectively infer ligand-receptor interactions in single-cell proteomics. The bulk data model has demonstrated therapeutic usefulness in cancer research.
