iTALK: an R Package to Characterize and Illustrate Intercellular Communication

Abstract

Crosstalk between tumor cells and other cells within the tumor microenvironment (TME) plays a crucial role in tumor progression, metastases, and therapy resistance. We present iTALK, a computational approach to characterize and illustrate intercellular communication signals in the multicellular tumor ecosystem using single-cell RNA sequencing data. iTALK can in principle be used to dissect the complexity, diversity, and dynamics of cell-cell communication from a wide range of cellular processes.The TME has emerged as a key modulator of tumor progression, immune evasion, and emergence of the anti-tumor therapy resistance mechanisms 1, 2 . The TME includes a diversity of cell types such as tumor cells, a heterogeneous group of immune cells, and the nonimmune stromal components. Tumor cells orchestrate and interact dynamically with these non-tumor components, and the crosstalk between them is thought to provide key signals that can direct and promote tumor cell growth and migration. Through this intercellular communication, tumor cells can elicit profound phenotypic changes in other TME cells such as tumor-associated fibroblasts, macrophages and T cells, and reprogram the TME, in order to escape from immune surveillance to facilitate survival. Therefore, a better understanding of the cell-cell communication signals may help identify novel modulating therapeutic strategies for better patient advantage. However, this has been hampered by the lack of bioinformatics tools for efficient data analysis and visualization.Here, we present iTALK (identifying and illustrating alterations in intercellular signaling network; https://github.com/Coolgenome/iTALK), an open source R package designed to profile and visualize the ligand-receptor mediated intercellular cross-talk signals from singlecell RNA sequencing data (scRNA-seq) ( Fig. 1 and Online Methods). We demonstrated that iTALK can be successfully applied to scRNA-seq data to capture highly abundant ligandreceptor gene (or transcript) pairs, identify gains or losses of cellular interactions by comparative analysis, and track the dynamic changes of intercellular communication signals in longitudinal samples. Notably, functional annotation of ligand-receptor genes is automatically added with our curated iTALK ligand-receptor database, and the output can be visualized in different formats with our efficient data visualization tool, which is implemented as part of iTALK. This approach can be applied to data sets ranging from hundreds to hundreds of thousands of cells and is not limited by sequencing platforms. It is also noteworthy that, in addition to studying the TME, iTALK can also be applied to a wide range of biomedical research fields that involve cell-cell communication.

ICELLNET: a transcriptome-based framework to dissect intercellular communication

“…This provides a straightforward manner to infer communication when a match is identified between a ligand and a cognate receptor for two respective cell types. The largest of such databases 9 integrated over 2500 ligand-receptor pairs through literature mining and computational analysis, and has been exploited in multiple computational tools for predicting cell-to-cell communication 5,8,33,34 .…”

“…With ICELLNET, we have developed a fully manually curated database, combining biological relevance, ease of use, and experimental validation. Except for one study 5 , ICELLNET is the only database offering a classification of predicted interactions into biological families. Similar to CellPhoneDB 10 , ICELLNET takes into account the multiple subunits of ligands and receptors, by introducing logical rules for co-expression of protein subunits.…”

“…Existing tools infer communication between cells from scRNAseq datasets 4,5,7 . We have designed ICELLNET as a versatile tool, which can be applied to bulk cell profiles (Affymetrix or RNAseq) widely available in public databases, but also to fully documented scRNAseq datasets to infer communication between clusters or groups of cells.…”

“…The last point relates to the granularity that is structuring the biological information into families and subfamilies of functionally and structurally related molecules. We only found one tool that provides a classification into four families of communication molecules 5 , while suffering from other limitations in particular the lack of manual curation.…”

Integrating multimodal data sets into a mathematical framework to describe and predict therapeutic resistance in cancer

“…The development of resistance to chemotherapy is a major cause of treatment failure in cancer. Intratumoral heterogeneity and phenotypic plasticity play a significant role in therapeutic resistance (1,2) and individual cell measurements such as flow and mass cytometry (3) and scRNA-seq (4) have been used to capture and analyze this cell variability (5)(6)(7)(8). Although attempts have been made to extract dynamic information from scRNA-seq via pseudo-time (9) or RNA velocity approaches (10), these high-throughput "omics" approaches come from cancer cell populations at a single time point.…”