This project applies G-cross spatial statistics to characterize cell–cell interaction patterns in colorectal cancer (CRC) tissue. Using spatially resolved single-cell data, we quantify directional spatial relationships between cell types and evaluate how these interaction patterns associate with clinical outcomes.
The focus is on capturing spatial organization and interaction structure within the tumor microenvironment rather than cell-type abundance alone.
Data Description
Data type: Spatial imaging–derived single-cell maps (e.g., multiplex IHC / CODEX)
Resolution: Single-cell coordinates
Annotations include:
Cell type labels
-
Spatial (x, y) coordinates
-
Sample or image identifiers
-
Clinical outcome metadata
Raw imaging data are not included due to data-sharing restrictions.
- Spatial Preprocessing
Parsing cell-level spatial coordinates
Filtering low-quality cells and rare cell types
Aligning spatial data with clinical metadata
- G-Cross Spatial Statistics
Computation of directional G-cross functions between cell-type pairs
Quantification of spatial attraction and avoidance
Calculation of AUC-based interaction scores per sample
- Feature Engineering
Construction of spatial interaction feature matrices
Aggregation of interaction scores across tissue regions
Normalization across samples
- Outcome Association
Statistical association of spatial features with clinical outcomes
Identification of interaction signatures linked to prognosis
Exploratory modeling for outcome prediction
-
Which cell–cell spatial interactions characterize the CRC tumor microenvironment?
-
How do immune–tumor and stromal–immune interactions vary by outcome?
-
Do spatial interaction features provide information beyond cell-type abundance?
-
Which interaction patterns suggest immune exclusion or immune engagement?