Skip to contents

Annotate single cells with a C++ SciBet-compatible classifier

Source: R/RunSciBet.R
RunSciBet.Rd

Run a C++ implementation of the core SciBet classifier using a labeled reference Seurat object.

Usage

RunSciBet(
  srt_query,
  srt_ref,
  ref_group,
  query_group = NULL,
  query_assay = NULL,
  ref_assay = NULL,
  query_layer = "counts",
  ref_layer = "counts",
  features = NULL,
  nfeatures = 1000,
  additional_features_per_class = 0,
  input_transform = c("auto", "none", "expm1"),
  prefix = "scibet",
  store_model = TRUE,
  store_probabilities = FALSE,
  return_object = TRUE,
  verbose = TRUE
)

Arguments

srt_query

An object of class Seurat to be annotated with cell types.

srt_ref

An object of class Seurat storing the reference cells.

ref_group

Column name in the srt_ref metadata that represents the cell grouping.

query_group

Column name in the srt_query metadata that represents the cell grouping.

query_assay

Assay to be used for the query data. Default is the default assay of the srt_query object.

ref_assay

Assay to be used for the reference data. Default is the default assay of the srt_ref object.

query_layer, ref_layer

Assay layers used for query and reference.

features

Candidate features used by SciBet. If NULL, common genes between query and reference are used.

nfeatures

Number of entropy-test features selected from features.

additional_features_per_class

Additional high-expression features selected per reference class.

input_transform

How to transform extracted values before SciBet's internal log1p step. "auto" applies expm1 to "data" layers and no transform otherwise.

prefix

Prefix for metadata columns.

store_model

Whether to store the SciBet core and probabilities in srt_query@tools.

store_probabilities

Whether to store the full cell-by-class probability matrix. The default keeps annotations and maximum probability scores while avoiding a large result object on full-scale data.

return_object

Whether to return the annotated Seurat query object. If FALSE, return a lightweight list with annotations, scores, probabilities, and model components without copying srt_query.

verbose

Whether to print the message. Default is TRUE.

Value

A Seurat object with SciBet annotations in metadata and results in srt_query@tools[["SciBet"]].

Examples

data(panc8_sub)
keep <- panc8_sub$celltype %in% c("ductal", "alpha", "beta")
panc8_sub <- Seurat::NormalizeData(panc8_sub[, keep], verbose = FALSE)
reference <- panc8_sub[, panc8_sub$tech != "fluidigmc1"]
query <- panc8_sub[, panc8_sub$tech == "fluidigmc1"]
query <- RunSciBet(
  srt_query = query, srt_ref = reference,
  ref_group = "celltype", nfeatures = 200, verbose = FALSE
)
query <- RunStandardWorkflow(query, verbose = FALSE, linear_reduction_dims = 10)
#> ℹ [2026-10-11 23:47:14] Skip `log1p()` because `layer = data` is not "counts"
CellDimPlot(
  query,
  group.by = "scibet_annotation",
  label = TRUE, legend.position = "bottom"
)