Skip to contents

Compute SHAP Values for Spatial Transcriptomics Data

Source: R/compute_shap_spatial.R
compute_shap_spatial.Rd

This function computes SHAP (SHapley Additive exPlanations) values for spatial transcriptomics data using a predictive model trained on bulk RNA-seq data. It returns a long-format data frame with one row per sample-feature pair.

Usage

compute_shap_spatial(model, X_bulk, y_bulk, X_spatial)

Arguments

model

A fitted glmnet model object (e.g., from cv.glmnet) with named elements $model and $lambda.

X_bulk

A numeric matrix or data.frame of bulk RNA-seq gene expression (samples × features).

y_bulk

A numeric vector of phenotype labels corresponding to X_bulk.

X_spatial

A numeric matrix or data.frame of spatial transcriptomics expression (samples × features).

Value

A data.frame of SHAP values for each sample-feature pair with columns:

  • feature: Feature (gene) name.

  • phi: The SHAP value of that feature.

  • sample: The spatial sample identifier.

  • feature.value: Original feature expression value.

Details

  • Normalizes X_bulk by row-wise sum to ensure comparability across samples.

  • Replaces NA and NaN in X_spatial with 0 to ensure model stability.

  • Ensures feature alignment and column order consistency between bulk and spatial datasets.

  • Computes SHAP values using the iml::Shapley class for each spatial unit.

Author

Bin Duan

Examples

if (FALSE) { # \dontrun{

data("osmFISH_metadata_cellType")
data("osmFISH_bulk_decon")
data("osmFISH_bulk_pheno")
data("osmFISH_bulk_coordinate")

PhenoResult <- SpatialPhenoMap(
  bulk_decon = osmFISH_bulk_decon,
  bulk_pheno = osmFISH_bulk_pheno,
  family = "binomial",
  coord = osmFISH_bulk_coordinate,
  resolution = "single_cell",
  sample_information_cellType = osmFISH_metadata_cellType
)

pred_result <- PhenoResult$pred_score
phenoPlus <- row.names(pred_result[pred_result$label %in% "phenotype+", ])
model <- PhenoResult$model
X <- as.data.frame(PhenoResult$cell_type_distribution[phenoPlus, ])

# This step took a very long time
shap_test_plus <- compute_shap_spatial(
  model = model,
  X_bulk = as.data.frame(osmFISH_bulk_decon),
  y_bulk = osmFISH_bulk_pheno,
  X_spatial = X)
} # }