CytoFlow: A Novel Computational Method to Construct Signal Transduction Networks at Single-Cell Resolution Based on Flow Networks

Abstract

The signal transduction network is essential for eukaryotic cellular communication and response to environmental signals, with disruptions leading to various diseases. As our understanding of intracellular signaling expands, the demand for computational tools that can efficiently synthesize this information into comprehensive networks grows. To meet this need, we developed CytoFlow, a novel computational model designed to construct cell-type-specific signal transduction networks by leveraging protein-protein interaction and single-cell transcriptomics data. Specifically, it models the network from receptors to transcription factors as a penalized flow network, optimized using linear programming techniques. We validated CytoFlow's superior precision against existing methods in reconstructing the yeast mitogen-activated protein kinase pathway, and demonstrated its ability to identify cell-type-specific signaling patterns in the human prefrontal cortex and peripheral blood mononuclear cells. In summary, CytoFlow offers a precise and cost-effective solution for constructing detailed signal transduction networks, advancing our capacity to understand and analyze cellular signaling processes.