Robustly and comprehensively deciphering cell-cell communication from distinct single-cell omics with a unified framework combining LIANA and Tensor-cell2cell
Erick Armingol (PhD candidate), University of California, San Diego, Advisor: Nathan E. Lewis
Daniel Dimitrov (PhD candidate) Heidelberg University, Advisor: Julio Saez-Rodriguez
Hratch Baghdassarian (PhD candidate), University of California, San Diego, Advisor: Nathan E. Lewis
Problem or question being addressed:
Cell-cell communication (CCC) coordinates higher-order biological functions in multicellular organisms (Almet et al. 2021; Armingol et al. 2021), dictating phenotypes in response to different contexts such as disease state, spatial location, and organismal life stage. In recent years, many tools have been developed to leverage single-cell and spatial transcriptomics to understand CCC driving various biological processes (Armingol et al. 2021). While each tool contributes unique and valuable features, many are tool-specific and challenging to integrate, with a plethora of scoring functions and databases housing prior knowledge (Armingol et al. 2021; Dimitrov et al. 2022). Moreover, most tools do not account for the relationships of CCC events across different contexts (Shakiba et al. 2021), either disregarding context altogether by analyzing samples individually or being limited to pairwise comparisons. Thus, as the transcriptomics field matures, and moves away from steady-state atlases, toward disease-related, multi-context comparisons, the interest to robustly disentangle multicellularity continues to increase (Petukhov et al. 2022).
Recently, we have developed two relevant tools. LIANA is a computational framework that runs any combination of available methods and ligand-receptor resources to infer CCC (Dimitrov et al. 2022). Tensor-cell2cell is a method that uncovers distinctive context-driven patterns of CCC across multiple samples simultaneously (Armingol et al. 2022). Our aim is to combine and expand our tools to create a generalized pipeline of CCC analysis that can handle multiple single-cell omics types (e.g., RNAseq or spatial), infer communication across multiple biological layers (e.g., protein- and metabolite-mediated), deconvolute multiple contexts simultaneously to distinguish the intercellular events driving phenotypic changes.
Rationale for your approach:
The rapid emergence of single-cell and spatially-resolved omics data has led to an explosion in methods development enabling researchers to decipher cell-cell communication (CCC) (Armingol et al. 2021). While each method comes with its own assumptions and provides unique insights, each is limited in its ability to analyze multiple contexts (i.e., samples) simultaneously. Furthermore many of their proposed developments are tool-specific and often incompatible with other tools for downstream integration. In addition, in recent work we saw that the choice of method and resource can notably affect the results, and thereby the biological interpretation and insights (Dimitrov et al. 2022). Thus, our goal is to consolidate existing methods and create an unified framework for robustly inferring cell-cell communication from different omics technologies, while simultaneously considering multiple conditions or cellular contexts. For this purpose, we are going to integrate and leverage LIANA (Dimitrov et al. 2022) and Tensor-cell2cell (Armingol et al. 2022).
Details of suggested approach:
Tensor-cell2cell can take the output of most other CCC tools, which have been generalized by LIANA, as its input (Fig. 1a). Together, LIANA and Tensor-cell2cell unify existing approaches into a general framework that enables researchers to select their preferred CCC resources (Fig. 1b), comparing the outputs of different methods, generating consensus results, and evaluating multiple samples or conditions simultaneously. By integrating the two tools, these analyses can be robustly conducted on any number of samples to assess key CCC mediators (Fig. 1c) without the additional complications of installing separate tools or reconciling discrepancies between them.
We will adapt our tools to leverage spatial omics data by defining cellular neighborhoods or niches (Tanevski et al. 2022), enabling us to infer relevant communication across different tissue locations. Next, in contrast to cell-cell communication mediated by proteins, generally estimated using expression levels of the genes encoding the interacting partners, communication mediated by metabolites can be inferred using enzyme expression as an indirect proxy of the production or consumption of metabolites. Thus, the resulting communication scores of proteins and metabolites may not be directly comparable. Here, we will harmonize protein and metabolite CCC inference to enable the extraction of concerted context-driven communication patterns coming from both types of molecules simultaneously. Finally, we will harmonize our tools with a vast general biological knowledge database (Türei et al. 2021), to improve interpretability of CCC predictions with downstream intracellular signaling information (Fig. 2).
How it will affect the broader field:
While CCC is a fast-growing field, the disparity between methods complicates the creation of user-friendly workflows that can incorporate distinct datasets and leaves gaps between otherwise apparent connections. Such connections include the integration of metabolite-mediated (Zheng et al. 2022; Garcia-Alonso et al. 2022) with protein-mediated scoring; combining disparate scoring functions (Armingol et al. 2021; Dimitrov et al. 2022) with context-dependent extraction of patterns (Armingol et al. 2022), and conducting downstream analysis developed across various tools (Fig. 2). As such, our unified framework will enable researchers from different fields to robustly and comprehensively characterize cell-cell communication in their dataset(s) of interest in a user-friendly manner.
Figure 1. Overview of the workflow for studying cell-cell communication through LIANA and Tensor-cell2cell. (a) General integration of LIANA and Tensor-cell2cell for running analysis across multiple samples, conditions or contexts. (b) Main inputs, steps, resources and options available to tune the inference of active ligand-receptor pairs in sender-receiver cell pairs. (c) Integration and simultaneous analysis of multiple samples to extract context-dependent patterns of cell-cell communication.
Figure 2. Downstream analyses of the cell-cell communication inference. (a) Visualization of active ligand-receptor pairs in sender-receiver cell pairs with a dot plot. (b) Circos or chord plot to summarize main ligand-receptor pairs used by cell pairs. (c) Network visualization of the overall cell-cell communication between cells. (d) Heatmap visualization of the number of ligand-receptor pairs used by each pair of cells. (e) Clustering of multiple samples given their context-dependent patterns of cell-cell communication. (f) Box plots and statistical analysis to compare groups of samples given their context-dependent patterns. (g) Enrichment analysis of the pathways associated with active ligand-receptor pairs from results of single- or multi-context CCC predictions. (h) Training of machine learning models from the CCC results.
References:
Almet, Axel A., Zixuan Cang, Suoqin Jin, and Qing Nie. 2021. “The Landscape of Cell–cell Communication through Single-Cell Transcriptomics.” Current Opinion in Systems Biology 26 (June): 12–23.
Armingol, Erick, Hratch M. Baghdassarian, Cameron Martino, Araceli Perez-Lopez, Caitlin Aamodt, Rob Knight, and Nathan E. Lewis. 2022. “Context-Aware Deconvolution of Cell-Cell Communication with Tensor-cell2cell.” Nature Communications 13 (1): 3665.
Armingol, Erick, Adam Officer, Olivier Harismendy, and Nathan E. Lewis. 2021. “Deciphering Cell-Cell Interactions and Communication from Gene Expression.” Nature Reviews. Genetics 22 (2): 71–88.
Dimitrov, Daniel, Dénes Türei, Martin Garrido-Rodriguez, Paul L. Burmedi, James S. Nagai, Charlotte Boys, Ricardo O. Ramirez Flores, et al. 2022. “Comparison of Methods and Resources for Cell-Cell Communication Inference from Single-Cell RNA-Seq Data.” Nature Communications 13 (1): 3224.
Garcia-Alonso, Luz, Valentina Lorenzi, Cecilia Icoresi Mazzeo, João Pedro Alves-Lopes, Kenny Roberts, Carmen Sancho-Serra, Justin Engelbert, et al. 2022. “Single-Cell Roadmap of Human Gonadal Development.” Nature 607 (7919): 540–47.
Petukhov, Viktor, Anna Igolkina, Rasmus Rydbirk, Shenglin Mei, Lars Christoffersen, Konstantin Khodosevich, and Peter V. Kharchenko. 2022. “Case-Control Analysis of Single-Cell RNA-Seq Studies.” bioRxiv. https://doi.org/10.1101/2022.03.15.484475.
Shakiba, Nika, Ross D. Jones, Ron Weiss, and Domitilla Del Vecchio. 2021. “Context-Aware Synthetic Biology by Controller Design: Engineering the Mammalian Cell.” Cell Systems 12 (6): 561–92.
Tanevski, Jovan, Ricardo Omar Ramirez Flores, Attila Gabor, Denis Schapiro, and Julio Saez-Rodriguez. 2022. “Explainable Multiview Framework for Dissecting Spatial Relationships from Highly Multiplexed Data.” Genome Biology 23 (1): 97.
Türei, Dénes, Alberto Valdeolivas, Lejla Gul, Nicolàs Palacio-Escat, Michal Klein, Olga Ivanova, Márton Ölbei, et al. 2021. “Integrated Intra- and Intercellular Signaling Knowledge for Multicellular Omics Analysis.” Molecular Systems Biology 17 (3): e9923.
Zheng, Rongbin, Yang Zhang, Tadataka Tsuji, Lili Zhang, Yu-Hua Tseng, and Kaifu Chen. 2022. “MEBOCOST: Metabolic Cell-Cell Communication Modeling by Single Cell Transcriptome.” bioRxiv. https://doi.org/10.1101/2022.05.30.494067.