Calculate Full Uncertainty Suite with Sample Grouping
Arguments
- prob_mat
The [Cells x Types] matrix from your kNN or randomforest
- spe
The SpatialExperiment object.
- sample_col
The column identifying different images/samples.
- k_spatial
Neighbors for the spatial discordance check (default 15). from training) to rbind with prob_mat before processing. Both must share the same column names.
Value
A data frame with one row per cell and columns for normalized entropy, Gini impurity, margin uncertainty, and spatial discordance.
Examples
data("cytoGateR_example", package = "cytoGateR")
image_ids <- SummarizedExperiment::colData(cytoGateR_example)$image_name
spe <- cytoGateR_example[, image_ids == image_ids[1L]]
prob_cols <- grep(
"^KNN_P_",
names(SummarizedExperiment::colData(spe)),
value = TRUE
)
prob_mat <- as.matrix(SummarizedExperiment::colData(spe)[, prob_cols])
rownames(prob_mat) <- colnames(spe)
colnames(prob_mat) <- sub("^KNN_P_", "", colnames(prob_mat))
uncertainty <- calculate_uncertainty(
prob_mat,
spe,
sample_col = "sample_id",
k_spatial = 3
)
head(uncertainty)
#> cell_id entropy gini_impurity margin_uncertainty spatial_discordance
#> 12_1 12_1 0 0 0 1.0000000
#> 12_2 12_2 0 0 0 0.3333333
#> 12_3 12_3 0 0 0 0.0000000
#> 12_4 12_4 0 0 0 0.6666667
#> 12_5 12_5 0 0 0 0.6666667
#> 12_6 12_6 0 0 0 0.6666667