hoodscanR: profiling single-cell neighborhoods in spatial transcriptomics data
TL;DR
The authors developed hoodscanR, a Bioconductor package that generates probabilistic, cell-level neighborhood profiles for spatial transcriptomics data to characterize mixed tissue environments.
Problem / question
Existing methods for identifying cellular neighborhoods in spatial data rely on hard assignments, failing to capture the partial membership and mixed environments of individual cells at the single-cell level.
Methods
The researchers developed hoodscanR, an R/Bioconductor package that calculates cell-level neighborhood probability profiles based on annotation-defined neighborhoods, and applied it to spatial transcriptomics datasets of breast and lung cancer.
Key findings
hoodscanR successfully generated probabilistic representations of cellular neighborhoods, allowing cells to have partial membership across multiple environments. When applied to breast and lung cancer datasets, the tool enabled the characterization of mixed tissue environments and successfully identified transcriptional changes in tumor cells located in distinct spatial neighborhoods. It also facilitated downstream tasks like uncertainty assessment and neighborhood-aware differential expression.
Why it matters
Providing a probabilistic rather than hard-assignment approach allows researchers to more accurately model complex, mixed tissue environments and discover spatially driven gene expression changes in diseases like cancer.
Limitations
The provided text does not explicitly mention any limitations, caveats, or constraints of the hoodscanR package.
Takeaway
hoodscanR improves spatial transcriptomics analysis by assigning cells probabilistic memberships to multiple neighborhoods, enabling finer-grained analysis of mixed tissue environments like tumors.