Spatial Deconvolution and Interaction Analysis in Complex Tissue Models

Cancer tissue is intrinsically heterogeneous, with tumour cells, immune infiltrates, and stroma forming a complex and highly interactive spatial architecture. Spatial transcriptomics provides direct molecular readout of these ecosystems within their native histological context. However, every measured spatial profile is a complex mixture. Each profile blends multiple cell types with overlapping intracellular processes, dynamic phenotypic states, and — in xenograft models — transcripts from two distinct genomes. Recovering separable, biologically meaningful signals from these mixed profiles is essential for understanding tumour-microenvironment interactions and translating spatial data into mechanism. Independent Component Analysis (ICA) offers a reference-free approach to disentangle these overlapping sources; stability is achieved through consICA consensus approach. We extend the consICA framework to spatial transcriptomics for robust signal deconvolution and host-tumour interaction analysis in complex tissue models.Presented at the European Bioconductor Conference (EuroBioC2026), Turku, June 2026.