Single-cell and spatial multi-omics for studying cell competition in the tumor microenvironment

Cell competition is a conserved fitness-surveillance process, first described in Drosophila epithelia, by which cells compare relative fitness with their neighbors so that less-fit “loser” cells are eliminated, extruded or forced to differentiate while fitter “winner” cells expand. In the tumor microenvironment (TME), this logic appears to be co-opted by transformed clones that eliminate normal epithelium, outcompete neighboring subclones and reshape stromal and immune compartments, making cell competition a potentially underappreciated contributor to cancer clonal evolution, immune escape and therapy resistance. Because competition is defined by a comparison between adjacent cells and by a demonstrated change in the relative representation of one population, it is difficult to resolve with bulk sequencing, which averages across precisely the heterogeneity that defines it. The convergence of single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, spatial and single-cell proteomics, artificial intelligence (AI) and network-inference bioinformatics now allows competition-relevant cellular states, interfaces and signaling relationships to be measured at the resolution the problem requires. We emphasize at the outset that these technologies are, in almost all current applications, hypothesis-generating: they identify candidate competitive interfaces and putative winner and loser states from static measurements, and they do not by themselves demonstrate that competition has occurred. This review synthesizes advances across these domains; proposes an explicit three-tier evidence framework separating direct experimental evidence of cell competition from competition-relevant associations and from computational predictions; distinguishes cell competition from related processes such as clonal selection, immune editing and differential proliferation; and defines a putative cell competition network as a falsifiable construct whose edges include ligand-receptor, spatial-adjacency, metabolic and mechanical relations. We further outline a conceptual, context-dependent Cell Competition Score (CCS) as a testable specification rather than a biomarker-ready instrument, and set out the benchmarking, causal-inference and prospective-validation requirements that stand between current capability and any clinical application.