Cell interactions refers to the physical contact of cells in a 2D space. In Enable Medicine insight reports, we typically use this term to refer to pairwise contact analysis—analyzing whether pairs of cell types interact more than we would expect based solely on the frequency of the cells in the tissue. Interpreting these interactions is important for understanding signaling processes, immune responses, and other cellular activity.

Our insight reports are built using SpatialMap, which has a number of methods for computing and analyzing pairwise contacts. These methods are described in detail in our user manual.

For our insights reports, our standard approach is to use a Delaunay Triangulation-based definition of pairwise contacts between cells.

After identifying spatial contacts, we then calculate enrichment scores for the interactions observed between each possible pair of cell types, accounting for both the frequency of each of the cell types and for the graph structure of the interconnections between each cell in that sample. We use one of the three methods described here: log-odds of interaction enrichment, hypergeometric test, or a permutation based method, depending on the dataset in question:

• For datasets where each biological sample is represented by a single, spatially-contiguous section of tissue, we will calculate a sample-level statistical test of enrichment using the hypergeometric test, and a parametric method of calculating the log-odds of enrichment.
• For datasets where at least one biological sample is represented by more than one separate, spatially discontinuous areas of tissue, we will calculate the enrichment test using the permutation based approach, and the log-odds of enrichment will be derived by comparison to the permuted null distribution.

If a metadata trait specifying cohort definitions is provided, the result of a statistical comparison of the log-odds of enrichment between groups will also be reported for each possible pair of cell interactions:

• For cohort definitions with two groups, the result of a two sample t-test will be reported.
• For cohort definitions with more than two groups, the result of a one-way ANOVA will be reported, with a Tukey’s post-hoc test performed to identify significant pairwise differences following significant ANOVA results.
• All reported p-values will be corrected for multiple hypotheses using a Benjamini-Hochberg correction.