LabelHarmonization: ELeFHAnt Label Harmonization

In praneet1988/ELeFHAnt: Ensemble Learning for Harmonization and Annotation of Single Cells

ELeFHAnt Label Harmonization

Description

Label Harmonization is a function to harmonize cell labels (celltypes) across single cell datasets. It requires a list of processed Seurat Objects or a integrated seurat object.

Usage

LabelHarmonization(
  seurat.objects = c(),
  perform_integration = FALSE,
  integrated.atlas = NULL,
  downsample = TRUE,
  downsample_to = 1000,
  npcs = 30,
  resolution = 0.8,
  validatePredictions = TRUE,
  selectanchorfeatures = 2000,
  ntree = 500,
  k.anchor = 5,
  k.filter = 200,
  k.score = 30,
  dims = 1:30,
  species = NULL,
  tissue = NULL,
  annotationCol = "Celltypes"
)

Arguments

seurat.objects	a list of processed seurat objects (please set Default Assay to "RNA") with Celltypes column in their respective meta.data to perform integration on
perform_integration	logical Indicator (TRUE or FALSE) to perform integration using list of seurat.objects
integrated.atlas	an integrated seurat object with CellTypes and seurat_clusters column in meta.data. Required if perform_integration = FALSE
downsample	logical Indicator (TRUE or FALSE) to downsample Seurat objects or integrated seurat object, enabling fast computation
downsample_to	a numerical value > 1 to downsample cells Default: 1000
npcs	number of principal components to compute after integration Default: 30
resolution	value of the resolution parameter, decides size of cell communities. Default: 0.8
validatePredictions	logical indicator (TRUE or FALSE) to asses predictions by deploying Gene set enrichment analysis
selectanchorfeatures	number of anchor features to use for integrating datasets (Default: 2000)
ntree	number of trees randomForest classifier should build (Default: 500)
k.anchor	How many neighbors (k) to use when picking anchors Default: 5
k.filter	How many neighbors (k) to use when filtering anchors Default: 200
k.score	How many neighbors (k) to use when scoring anchors Default: 30
dims	Which dimensions to use from the CCA to specify the neighbor search space Default: 1:30
species	human or mouse if validatePredictions = TRUE
tissue	please check human_tissues or mouse_tissues if validatePredictions = TRUE
annotationCol	Name of column in metdata which has Celltypes information Default: Celltypes. It has to be same across atlases you want to integrate or have integrated

Value

integrated seurat object with harmonized celltypes added to meta.data of the object

Author(s)

Praneet Chaturvedi & Konrad Thorner

praneet1988/ELeFHAnt documentation built on July 20, 2022, 7:29 p.m.