Abstract
Cell-cell communication (CCC) inference has become a routine task in single-cell data analysis. Many computational tools are developed for this purpose. However, the robustness of existing CCC methods remains underexplored. We develop a user-friendly tool, RobustCCC, to facilitate the robustness evaluation of CCC methods with respect to three perspectives, including replicated data, transcriptomic data noise and prior knowledge noise. RobustCCC currently integrates 14 state-of-the-art CCC methods and 6 simulated single-cell transcriptomics datasets to generate robustness evaluation reports in tabular form for easy interpretation. We find that these methods exhibit substantially different robustness performances using different simulation datasets, implying a strong impact of the input data on resulting CCC patterns. In summary, RobustCCC represents a scalable tool that can easily integrate more CCC methods, more single-cell datasets from different species (e.g., mouse and human) to provide guidance in selecting methods for identification of consistent and stable CCC patterns in tissue microenvironments. RobustCCC is freely available at https://github.com/GaoLabXDU/RobustCCC.