Image-based spatial transcriptomics enables the visualization of RNA molecules within intact tissues at subcellular resolution. However, a substantial proportion of transcripts often remains unassigned during cell segmentation. These unassigned RNAs (uRNAs) are typically excluded from downstream analyses, despite their potential biological relevance.

In this study, we explore the origins and biological relevance of uRNA. We systematically investigate uRNAs across 14 datasets from different tissue types generated with diverse technologies. To facilitate broader exploration of uRNAs, we developed Troutpy, an open-source Python package that offers tools for exploring uRNA, including the detection of undersegmentation events, identification of cell–cell contact zones, and analysis of extrasomatic RNAs.

By quantifying the contributions of various technical factors, including undersegmentation, imaging noise, and molecular diffusion, we identify both technical and non-technical sources of uRNAs. Notably, a significant portion of uRNAs exhibits non-random spatial patterns and local enrichment in specific tissue structures, particularly in regions associated with distinct extrasomatic features. We demonstrate these patterns to be reproducible across datasets and platforms, suggesting a biological origin. Finally, we demonstrate that uRNAs exhibit distinct spatial expression signatures, divergent from intracellular profiles, that align with specific cellular architectures, thereby offering a framework to investigate the extrasomatic distribution of RNA across cell types.

Our findings highlight the biological relevance of unassigned RNAs (uRNA) in spatial transcriptomics. By proposing strategies to incorporate uRNA into existing analyses, our study extends the application of spatial transcriptomics to investigate extrasomatic RNA, potentially key in tissue organization, and cell–cell interactions or migration, among other topics.