Ultra-precision deconvolution of spatial transcriptomics decodes immune heterogeneity and fate-defining programs in tissues

Elucidating the spatial organization and functional specialization of immune cells within complex tissues remains challenging. We present UCASpatial, an ultra-precision spatial transcriptomics deconvolution algorithm utilizing entropy-based weighting to accurately map cell subpopulations. Benchmarking confirms its superiority in identifying low-abundant cell subpopulations and distinguishing transcriptionally heterogeneous cell subpopulations. Applying UCASpatial to human colorectal cancer, we reveal that chromosome 20q gain in individual cancer clones orchestrates a T cell-excluded microenvironment, associated with HERV-H silencing and impaired type I interferon responses. In murine wound healing models, we reveal spatiotemporal dynamics distinguishing scarring from regenerative phenotypes. Specifically, we identify a pro-fibrotic community comprising Igfbp5+ chondrocytes, Cd36+ Gpnmb+ Il1b- macrophages, and Fmod+ fibroblasts in scarring-healing mice (C57BL/6). We further demonstrate that IL11-IL11RA signaling within this triad drives the pro-fibrotic community formation and limits regeneration. Together, UCASpatial serves as a versatile tool for deciphering fine-grained cellular landscapes and exploring intercellular mechanisms in complex and dynamic microenvironments. Accurately mapping diverse cell types in complex tissues remains a challenge for spatial transcriptomics. Here, the authors present UCASpatial, an entropy-weighted algorithm that enables ultra-precision deconvolution to resolve fine-grained immune landscapes and explore intercellular mechanisms.