TL;DR
This paper introduces OmniPath, a comprehensive resource that integrates over 100 databases to map both intracellular and intercellular signaling pathways. This unified knowledge base is crucial for analyzing complex multicellular biological systems, especially with the rise of single-cell omics data, and is made accessible through various computational tools.
What Problem Does This Paper Address?
Understanding how cells communicate and coordinate their activities is fundamental to comprehending complex biological processes, especially in multicellular organisms. While significant progress has been made in cataloging intracellular signaling pathways within individual cells, knowledge about how cells communicate with each other (intercellular signaling) has been fragmented, scattered across numerous resources, and often not linked to the intracellular processes they trigger. This gap makes it challenging to perform comprehensive analyses of multicellular systems, particularly when using high-throughput "omics" data (like transcriptomics) from individual cells. Researchers need a unified resource that connects the dots between what happens inside a cell and how it responds to signals from its neighbors.
Methods
The core of this work is the creation of OmniPath, a unified knowledge base built upon the existing OmniPath resource. The authors integrated data from over 100 diverse sources, encompassing:
- Molecular Interactions: Protein-protein interactions (PPI), ligand-receptor interactions, and enzyme-post-translational modification (PTM) relationships.
- Regulation: Transcriptional regulation (TF-gene) and miRNA-mRNA interactions.
- Complexes: Information on protein complexes.
- Annotations: Protein function, localization, and roles in diseases.
A key innovation is the development of a comprehensive set of intercellular communication roles for proteins, categorizing them into eight major and seventeen minor functional roles (e.g., ligand, receptor, extracellular matrix component, transporter). This classification was meticulously curated and cross-referenced with UniProt data.
The data is managed and made accessible through a Python module called pypath, which builds five integrated databases: Network, Enzyme-PTM, Complexes, Annotations, and Intercell. These databases are then exposed via a web service, an R/Bioconductor package (OmnipathR), a Python client, and a Cytoscape plugin, offering flexible access for computational and experimental scientists.
Datasets
OmniPath integrates data from over 103 original resources. These resources cover a wide spectrum of biological knowledge, including:
- Protein-protein interactions: 38 resources.
- Ligand-receptor interactions: 10 resources.
- Enzyme-PTM relationships: 11 resources.
- Protein complexes: 12 resources.
- Annotations (function, localization, etc.): 49 resources.
- Transcriptional regulation: 7 resources.
- miRNA-mRNA interactions: 5 resources.
- TF-miRNA interactions: 2 resources.
The paper also mentions using data from specific studies for case studies, such as RNA-Seq data from Calu3 cells (GSE147507) for SARS-CoV-2 analysis and single-cell RNA-Seq data from ulcerative colitis patients (Smillie et al., 2019).
Key Insights and Ideas
- Bridging the Intra- and Intercellular Divide: The most significant contribution is the seamless integration of knowledge about what happens inside a cell with how cells communicate with each other. This unified view is essential for understanding complex multicellular phenotypes.
- Standardized Intercellular Roles: Defining and systematically annotating specific roles for proteins in intercellular communication (ligand, receptor, etc.) provides a standardized framework for analysis, moving beyond just ligand-receptor pairs.
- Comprehensive Coverage: By integrating over 100 resources, OmniPath offers a much broader and deeper view of signaling networks than previously available, capturing interactions and annotations that might be missed by specialized databases.
- Flexibility and Accessibility: Providing data through multiple interfaces (web service, R/Python packages, Cytoscape plugin) makes OmniPath accessible to a wide range of users with different computational backgrounds and analysis preferences.
- Enabling Advanced Analysis: OmniPath serves as a crucial prior knowledge input for cutting-edge single-cell analysis tools like NicheNet and CellPhoneDB, facilitating mechanistic insights into complex diseases.
Key Results
- OmniPath integrates data from 103 resources into five distinct databases (Network, Enzyme-PTM, Complexes, Annotations, Intercell).
- The resource contains over 100,000 PPI interactions and over 14,000 directed ligand-receptor connections, significantly more than many specialized resources.
- 54,330 functional annotations related to intercellular communication roles were provided for 5,781 proteins.
- The resource includes 22,005 protein complexes.
- Case studies demonstrated OmniPath's utility in:
- Uncovering potential signaling mechanisms in SARS-CoV-2 infection, identifying pathways like JAK-STAT and MAPK, and suggesting potential drug targets (e.g., minocycline).
- Analyzing ulcerative colitis by mapping alterations in intercellular communication between cell types, revealing changes in Treg cell signaling and downstream effects.
- A comparison with other resources (Pathway Commons, Ramilowski) showed that OmniPath, particularly when used with its own annotations, provided a larger and more detailed network for analyzing intercellular interactions and their downstream effects.
Limitations and Future Work
- Homology Translation: While OmniPath includes human data, translation to rodent species (mouse and rat) is available but has varying success rates (81% for mouse, 31% for rat). Future work could expand this to more species.
- Context Specificity: The current knowledge in OmniPath is largely general and not specific to particular cell types or physiological conditions. The authors plan to incorporate more context information (e.g., cell type, condition) and use scores to prioritize interactions and paths that are more relevant to specific biological scenarios.
- Text Mining and Benchmarking: Future efforts will leverage text mining to extract more context and will involve benchmarking intercellular communication knowledge against experimental data to improve its reliability.
- Expanding Scope: The authors envision extending OmniPath to include pathogen-host interactions and microbiome-host interactions, further broadening its applicability.
Summary updated