Pathway Centric Analysis for single-cell RNA-seq and Spatial Transcriptomics Data with GSDensity

Abstract Advances in single-cell technology have enabled molecular cellular dissection of heterogeneous biospecimens at unprecedented scales and resolutions. Although cluster-centric approaches followed by gene-set analysis can reveal distinct cell types and states, they have limited power in dissecting and interpretating highly heterogenous, dynamically evolving data. Here, we present GSDensity, a graph-modeling approach that allows users to obtain pathway-centric interpretation and dissection of single-cell and spatial transcriptomics (ST) data without performing clustering. We show that GSDensity can not only accurately detect biologically distinct gene sets but also reveal novel cell-pathway associations that are ignored by existing methods. This is particularly evident in characterizing cancer cell states that are transcriptomically distinct but are driven by shared tumor-immune interaction mechanisms. Moreover, we show that GSDensity, combined with trajectory analysis can identify pathways that are active at various stages of mouse brain development. Finally, we show that GSDensity can identify spatially relevant pathways in mouse brains including those following a high-order organizational patterns in the ST data. We also created a pan-cancer pathway activity ST map, which revealed pathways spatially relevant and recurrently active across six different tumor types. GSDensity is available as an open-source R package and can be widely applied to single-cell and ST data generated by various technologies.