Microenvironment-aware transcriptome reconstruction in spatial transcriptomics

Abstract

Imaging-based spatial transcriptomics provides single-cell resolution, but remains limited to targeted gene panels, leaving much of the transcriptome and cellular-state variation unobserved. Existing approaches infer unmeasured genes primarily through transcriptional alignment between spatial and reference datasets, which effectively recover broad cell identities, but may not fully capture context-dependent variation within cell types. Here we show that local microenvironmental organization is associated with transcriptional variation beyond cell identity and provides complementary constraints for transcriptome reconstruction. Based on this observation, we present Emerge, a microenvironment-aware framework that integrates transcriptional identity with the local microenvironment to reconstruct transcriptome-scale expression. Across MERFISH and Xenium datasets spanning neural and tumor ecosystems, we show that Emerge improves recovery of structured transcriptional variation and reveals coordinated relationships among cellular states, ecological niches and intercellular communication programs that are only partially captured by targeted gene panels. Together, we demonstrate that Emerge extends transcriptome reconstruction from gene completion to context-aware recovery of cellular states. Current methods for spatial transcriptomics data analysis do not fully capture context-dependent variation within cell types. Here the authors present Emerge, which integrates cellular identity with tissue environments to reveal hidden cellular states and microenvironment-associated transcriptional programs.