scPredict: Predict cell classes from a new dataset using a trained model

In powellgenomicslab/scPred: scPred - accurate supervised method for cell-type classification from single-cell RNA-seq data

Description

Predicts cell classes for a new dataset based on trained model(s)

Usage

scPredict(
  new,
  reference,
  threshold = 0.55,
  max.iter.harmony = 20,
  recompute_alignment = TRUE,
  seed = 66
)

Arguments

new	A seurat object containing cells to be classified
reference	A Seurat object with trained model(s) using scPred or an scPred object
threshold	Threshold used for probabilities to classify cells into classes. All cells below this threshold value will be labels as "unassigned". In the case of binary classification (two cell tyoes), a threshold of 0.5 will force all cells to be classified to any of the two cell types. For multi-class classification, if there's no probability higher than the threshold associated to a cell type, this will be labelled as "unassigned"
max.iter.harmony	Maximum number of rounds to run Harmony. One round of Harmony involves one clustering and one correction step.
recompute_alignment	Recompute alignment? Useful if scPredict() has already been run
seed	Numeric seed for harmony

Value

A Seurat object with additional metadata columns with prediction probabilities associated to each class, a prediction column, indicating the classification based on the provided threshold and a generic_class column without "unassigned" labels. Additionally, two new reductions are returned:

• scpred: Aligned data using harmony

• scpred_projection: Raw projection using reference loadings

Author(s)

José Alquicira Hernández

powellgenomicslab/scPred documentation built on July 16, 2021, 12:14 a.m.