Predict unknown cells using a trained deep-learning model

Drop-in deep-learning counterpart to predict_unknown_with_randomforest. Applies a "cytoGateR_dl" model (from train_custom_dl) to cells currently labelled unknown_label (or NA) and fills in predictions above threshold.

Usage

predict_unknown_with_dl(
  spe,
  model,
  assay_name = "exprs",
  label_col = "custom_label",
  out_col = "soft_tree_label_filled",
  pred_col = "dl_pred",
  unknown_label = "Unknown",
  threshold = 0.5,
  unassigned_label = "Unassigned"
)

Arguments

spe

A SpatialExperiment or SingleCellExperiment.

model

A "cytoGateR_dl" object, as returned by train_custom_dl()$model.

assay_name

Character scalar naming the assay used as features (default "exprs").

label_col

Column in colData(spe) identifying which cells are unknown (default "custom_label").

out_col

Column name to store the filled-in labels (default "soft_tree_label_filled").

pred_col

Column name to store raw predictions for the unknown subset (default "dl_pred").

unknown_label

Character label treated as unlabeled (default "Unknown").

threshold

Minimum softmax probability required to accept a prediction (default 0.5).

unassigned_label

Label assigned when the top probability is below threshold (default "Unassigned").

Value

A named list with spe (updated) and prob_mat (probability matrix for the predicted-on cells only).

Examples

# \donttest{
if (requireNamespace("torch", quietly = TRUE) &&
    torch::torch_is_installed()) {
  data("cytoGateR_example", package = "cytoGateR")
  image_ids <- SummarizedExperiment::colData(cytoGateR_example)$image_name
  spe <- cytoGateR_example[, image_ids == image_ids[1L]]
  dl_fit <- train_custom_dl(
    spe,
    label_col = "core_group",
    unknown_label = "Unassigned",
    features = c("CD3e", "CD20", "CD4", "CD8a"),
    hidden_dims = 8,
    epochs = 2,
    batch_size = 256,
    cv_folds = 2,
    repeats = 1,
    agreement_thresh = 0,
    patience = 1,
    seed = 1
  )
  dl_prediction <- predict_unknown_with_dl(
    dl_fit$spe,
    dl_fit$model,
    label_col = "cleaned_core_label",
    unknown_label = "Unassigned",
    threshold = 0.5
  )
  head(SummarizedExperiment::colData(dl_prediction$spe)$dl_pred)
}
# }