Skip to contents

Run scMalignantFinder malignant cell identification

Source: R/RunscMalignantFinder.R
RunscMalignantFinder.Rd

Run the Python package scMalignantFinder on a Seurat or AnnData object and append malignant-cell predictions to Seurat metadata. The pretrained model files are not bundled with scop; provide a directory containing model.joblib and ordered_feature.tsv through pretrain_dir.

Usage

RunscMalignantFinder(
  srt = NULL,
  adata = NULL,
  h5ad = NULL,
  assay = "RNA",
  layer = "counts",
  cells = NULL,
  pretrain_dir = NULL,
  train_h5ad_path = NULL,
  feature_path = NULL,
  model_method = c("LogisticRegression", "RandomForest", "XGBoost"),
  norm_type = NULL,
  use_raw = FALSE,
  n_thread = 1,
  prefix = "",
  return_seurat = !is.null(srt),
  verbose = TRUE
)

Arguments

srt

A Seurat object.

adata

Optional Python AnnData object.

h5ad

Optional path to an .h5ad file.

assay

Assay used when srt is supplied. Default is "RNA".

layer

Layer used when srt is supplied. Default is "counts".

cells

Optional cells to run. If supplied with srt, results are appended to these cells and other cells receive NA.

pretrain_dir

Directory containing pretrained scMalignantFinder model files: model.joblib and ordered_feature.tsv.

train_h5ad_path

Optional training .h5ad file used when training a model from scratch.

feature_path

Optional feature file used when training from scratch.

model_method

Model used when training from scratch. One of "LogisticRegression", "RandomForest", or "XGBoost".

norm_type

Passed to scMalignantFinder. Use TRUE for raw counts that should be library-size normalized; use FALSE for already normalized input. If NULL, defaults to TRUE only for Seurat counts input.

use_raw

Whether to use adata.raw.X when available.

n_thread

Number of threads used by scMalignantFinder.

prefix

Optional prefix for output metadata columns. Default preserves the original scMalignantFinder column names.

return_seurat

Whether to return a Seurat object when srt is supplied. If FALSE, returns a data frame of predictions.

verbose

Whether to print the message. Default is TRUE.

Value

A Seurat object with scMalignantFinder_prediction and malignancy_probability added, or a data frame when return_seurat = FALSE.

References

Yu Q, Li YY, Chen Y. scMalignantFinder distinguishes malignant cells in single-cell and spatial transcriptomics by leveraging cancer signatures. Communications Biology. 2025.

See also

srt_to_adata, RunscMalignantRegion, RunscMalignantStates

Examples

if (FALSE) {
data(pancreas_sub)
pancreas_sub <- RunscMalignantFinder(
  pancreas_sub,
  assay = "RNA",
  layer = "counts",
  pretrain_dir = "path/to/pretrained_model"
)
CellDimPlot(pancreas_sub, group.by = "malignancy_probability")
}