scAlign: Run scAlign framework

In scAlign: An alignment and integration method for single cell genomics

Description

Main function for scAlign that runs encoder and decoder networks

Usage

scAlign(sce.object, options = scAlignOptions(), encoder.data,
  decoder.data = encoder.data, supervised = "none",
  run.encoder = TRUE, run.decoder = FALSE, log.dir = "./models/",
  log.results = FALSE, device = "CPU")

Arguments

sce.object	scAlign object.
options	Training options for scAlign.
encoder.data	Which data format to use for alignment.
decoder.data	Which data format to use for interpolation.
supervised	Run scAlign in supervised mode, requires labels.
run.encoder	Run scAlign alignment procedure.
run.decoder	Run scAlign projection through paired decoders.
log.dir	Location to save results.
log.results	Determines if results should be written to log.dir.
device	Specify hardware to use. May not work on all systems, manually set CUDA_VISIBLE_DEVICES if necessary.

Value

SingleCellExperiment

Examples

library(Seurat)
library(SingleCellExperiment)

## Input data, 1000 genes x 100 cells
data = matrix(sample.int(10000, 1000*100, TRUE), 1000, 100)
rownames(data) = paste0("gene", seq_len(1000))
colnames(data) = paste0("cell", seq_len(100))

age    = c(rep("young",50), rep("old",50))
labels = c(c(rep("type1",25), rep("type2",25)), c(rep("type1",25), rep("type2",25)))

## Build the SCE object for input to scAlign using Seurat preprocessing and variable gene selection
ctrlSCE <- SingleCellExperiment(
              assays = list(scale.data = data[,which(age == "young")]))

stimSCE <- SingleCellExperiment(
              assays = list(scale.data = data[,which(age == "old")]))

## Build the scAlign class object and compute PCs
scAlignHSC = scAlignCreateObject(sce.objects = list("YOUNG"=ctrlSCE,
                                                    "OLD"=stimSCE),
                                 labels = list(labels[which(age == "young")],
                                               labels[which(age == "old")]),
                                 pca.reduce = TRUE,
                                 pcs.compute = 50,
                                 cca.reduce = TRUE,
                                 ccs.compute = 15,
                                 project.name = "scAlign_Kowalcyzk_HSC")

## Run scAlign with high_var_genes
scAlignHSC = scAlign(scAlignHSC,
                   options=scAlignOptions(steps=1, log.every=1, norm=TRUE, early.stop=FALSE),
                   encoder.data="scale.data",
                   supervised='none',
                   run.encoder=TRUE,
                   run.decoder=FALSE,
                   log.dir=file.path('~/models','gene_input'),
                   device="CPU")

scAlign documentation built on April 28, 2020, 6:10 p.m.