Patient-Specific Cell Communication Networks Associate With Disease Progression in Cancer

Abstract

The maintenance and function of tissues in health and disease depends on cell–cell communication. This work shows how high-level features, representing cell–cell communication, can be defined and used to associate certain signaling “axes” with clinical outcomes. We generated a scaffold of cell–cell interactions and defined a probabilistic method for creating per-patient weighted graphs based on gene expression and cell deconvolution results. With this method, we generated over 9,000 graphs for The Cancer Genome Atlas (TCGA) patient samples, each representing likely channels of intercellular communication in the tumor microenvironment (TME). It was shown that cell–cell edges were strongly associated with disease severity and progression, in terms of survival time and tumor stage. Within individual tumor types, there are predominant cell types, and the collection of associated edges were found to be predictive of clinical phenotypes. Additionally, genes associated with differentially weighted edges were enriched in Gene Ontology terms associated with tissue structure and immune response. Code, data, and notebooks are provided to enable the application of this method to any expression dataset (https://github.com/IlyaLab/Pan-Cancer-Cell-Cell-Comm-Net).

Mathematically mapping the network of cells in the tumor microenvironment

“…A possible explanation for this in the context of the TME, is the observation that ligand-receptor interactions are not a one-way street. When a ligand-receptor interaction occurs, often this elicits a reaction in both the ligand cell-type, and the receptor cell-type 10 , partially masking the directionality of the interaction.…”

“…An unbiased approach to do this consists in the modeling of the TME as a cell-cell communication network, which can be inferred typically from RNA sequencing (RNAseq) data using statistical inference methods or machine learning techniques 6 . Several studies have shown the value of using the reconstructed cell-cell communication networks to study the role of cell-cell communication in the TME [7][8][9][10][11][12] . However, existing techniques have several drawbacks.…”

“…This provides "higher resolution" modeling, but is more complicated to apply in specific use cases, since scRNA-seq data itself has some technical limitations: higher uncertainty, drop-outs, and limited clinical applicability due to its higher costs and difficulties in sample preparation 14 . A recent approach has been proposed that combines bulk RNA-seq data with probabilistic techniques to reconstruct cell-cell interaction networks for individual patients 10 . However, this method builds a network on the level of cell-types that provides only a mean-field approximation to the actual cell-cell interaction network without mathematical guarantees on how well this approximation fits the data.…”

Technological and computational advances driving high-throughput oncology

“…Although more challenging, reconstructing cell-cell interactions from bulk data has the important advantage of being applicable to large available cohorts of tumor samples, with more direct clinical implications. Great advancements in this direction have recently been published, showing the potential of such methods in predicting clinical outcomes [73,74] and response to immunotherapy with checkpoint inhibitors [75,76].…”

Analyzing network diversity of cell–cell interactions in COVID-19 using single-cell transcriptomics

“…The majority of CCI inference methods are designed to analyze CCI for a single condition. As characterizing differences in CCI activity across multiple conditions—for example, between healthy and diseased tissue—helps elucidate the diverse mechanisms of CCI-mediated responses to disease, more so than analyzing condition-specific CCI networks in isolation, recent efforts have focused on developing methodologies to systematically study the differences in CCI between multiple conditions ( Xiong et al, 2019 ; Gibbs et al, 2021 ; Wang et al, 2022 ). However, the majority of previous CCI analyses that analyze differences between conditions tend to analyze CCI networks inferred from a single scRNA-seq sample from each condition or analyze CCI networks constructed by aggregating several samples.…”