Decoding Spatial Transcriptomics: Computational Frameworks for Subcellular Spatial Patterns and Contact Mediated Signaling
Author(s)
Kumar, Anurendra
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Spatial transcriptomics has revolutionized our ability to probe the spatial organization of gene expression, enabling a deeper understanding of cellular function, cell-cell communication, and tissue architecture. By bridging molecular detail with spatial context, these technologies have opened new avenues in the study of development, disease, and cellular heterogeneity. However, realizing their full potential requires computational methods that can extract meaningful insights from both subcellular and intercellular spatial patterns.
This thesis presents two complementary computational frameworks that address critical analytical gaps in spatial transcriptomics: one focusing on subcellular spatial organization, and the other on contact-mediated intercellular communication.
The first framework, InSTAnT (Intracellular Spatial Transcriptomic Analysis Toolkit), quantifies subcellular transcript localization and uncovers robust RNA co-localization patterns, an underexplored but biologically important dimension of gene regulation. InSTAnT addresses the lack of a universal coordinate system across cells and mitigates gene expression confounders to identify statistically significant spatial signatures. Applied across diverse datasets spanning different species, brain regions, and technologies, InSTAnT reveals consistent and biologically meaningful subcellular organization patterns, offering a principled foundation for investigating RNA localization as a regulatory mechanism.
The second framework developed in this thesis, called CellWHISPER (Workflow for HIgh-precision Spatial Proximity-mediated cEll-cell inteRactions), extends spatial analysis to the tissue level by modeling contact-mediated (e.g., gap junction) cell-cell communication. Existing methods often overlook spatial confounders, resulting in inflated false-positive interactions. CellWHISPER introduces the concept of a "whisper network" to map signaling between spatially adjacent cells and performs rigorous statistical testing to learn inter-cellular communication mechanisms; it also incorporates an analytical null model for computational scalibility. We develop a novel latent variable model for cell-type and gene-pair interaction analysis. Differential analysis on Alzheimer’s disease yields AD associated gap junction communication.
Together, InSTAnT and CellWHISPER represent an integrated, multi-scale strategy for interpreting spatial transcriptomic data, from the subcellular localization of individual transcripts to the collective communication patterns across cell populations. By developing statistically rigorous tools for spatial omics, this thesis contributes to a more spatially grounded and mechanistic understanding of cellular behavior in health and disease.
Sponsor
Date
2025-08-06
Extent
Resource Type
Text
Resource Subtype
Dissertation