geneSCOPE: gene spatial co-occurrence of pairwise expression

Spatial transcriptomics captures context-dependent gene expression; however, existing workflows do not consistently account for measurement scale and often rely on a user-defined spatial neighbor graph, making results sensitive to this choice and limiting cross-study comparability. We present geneSCOPE (gene Spatial Co-Occurrence of Pairwise Expression), a framework that integrates ecology-inspired spatial statistics with network analysis to explicitly capture measurement scale and spatial information. In this framework, molecules are binned on a grid with a width selected near the mode of the per-gene unit-invariant knee distribution derived from Morisita's ${I}_{\delta }$-width curves. Pairwise adjacency-weighted spatial associations are quantified using Lee's L. Then, a spatial gene network is assembled, and gene modules are identified via consensus clustering. Cell-cell interactions among various cell types are identified based on high Lee's L with low cell-level co-expression (Pearson's $r$). When applied to transcriptome data derived from human colorectal cancer and lymph node Xenium tissue sections (N = 3 and 1, respectively), geneSCOPE recovered spatial gene modules that mapped to microanatomical compartments such as invasive margins, luminal epithelium, fibroblast-rich territories, and germinal-center subdomains. Further, it highlighted intercellular neighborhood patterns at tumor-stroma interfaces characterized by the co-occurrence of leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5)-marked stem-like tumor programs and complement component 3 (C3)-centered fibroblast/complement-associated niches. Using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database as an external reference for benchmarking, geneSCOPE showed the highest concordance with known interacting gene pairs among the compared methods. In conclusion, geneSCOPE provides a scalable, interpretable, and cross-study comparable framework for gene-centric spatial analysis.