SpaLinker identifies phenotype-associated spatial tumor microenvironment features by integrating bulk and spatial sequencing data.

Abstract

The emergence of spatial transcriptomics (ST) technology offers unprecedented opportunities to elucidate the complexity and heterogeneity of the tumor microenvironment (TME). However, quantitatively linking spatially resolved features with clinical phenotypes remains challenging due to the scarcity of clinical annotations of spatial sequencing samples. Herein, we introduce SpaLinker, an innovative integrated framework that utilizes ST data to decipher spatially resolved TMEs at molecular, cellular, and tissue structure levels. Specifically, it assesses the prognostic significance of spatially defined features by integrating well-accumulated bulk RNA sequencing (RNA-seq) data, using a phenotype-driven computational framework. Applying SpaLinker to diverse tumor ST datasets demonstrated its utility and effectiveness in recognizing spatial architectures, including tertiary lymphoid structures and tumor-normal interfaces, and in establishing links to distinct clinical outcomes. Overall, this study presents a valuable and comprehensive pan-cancer analytical platform to de novo identify phenotype-associated spatial TME features, significantly enhancing the clinical utility of spatial sequencing technology.