Abstract
Spatial omics technologies enable the simultaneous measurement of molecular features within their spatial context, providing unprecedented insights into cellular organization and tissue architecture. The advent of single-cell resolution platforms has further enhanced our ability to uncover microenvironment-dependent cell states. While numerous computational methods have been developed for spatial omics analysis, most focus on spatial domain detection rather than on resolving cell identities. Traditional single-cell clustering methods rely solely on intrinsic molecular features to determine cell identities, missing the impact of the local microenvironment on cell states. To address this gap, we present MEcell, a parameter-free method that explicitly incorporates spatial context and automatically adjusts its contribution to improve cell identity modeling. Applied to 90 simulated and 7 real datasets spanning multiple spatial transcriptomics platforms and tissue types, including MERFISH/Vizgen, Xenium, CosMx, Visium HD, Slide-seqV2, and open-ST, MEcell consistently outperformed existing methods in accurately inferring cell identities. These results highlight the critical role of microenvironment in defining cell identity and demonstrate the power of MEcell for capturing spatially informed cellular heterogeneity.