Seurat (version 2.3.4)

CalcAlignmentMetric: Calculate an alignment score

Description

Calculates an alignment score to determine how well aligned two (or more) groups have been aligned. We first split the data into groups based on the grouping.var provided and randomly downsample all groups to have as many cells as in the smallest group. We then construct a nearest neighbor graph and ask for each cell, how many of its neighbors have the same group identity as it does. We then take the average over all cells, compare it to the expected value for perfectly mixed neighborhoods, and scale it to range from 0 to 1.

Usage

CalcAlignmentMetric(object, reduction.use = "cca.aligned", dims.use,
  grouping.var, nn, nn.eps = 0)

Arguments

object

Seurat object

reduction.use

Stored dimensional reduction on which to build NN graph. Usually going to be cca.aligned.

dims.use

Dimensions to use in building the NN graph

grouping.var

Grouping variable used in the alignment.

nn

Number of neighbors to calculate in the NN graph construction

nn.eps

Error bound when performing nearest neighbor seach using RANN; default of 0.0 implies exact nearest neighbor search

Details

xbar is the average number of neighbors belonging to any cells' same group, N is the number of groups in the given grouping.var, k is the number of neighbors in the KNN graph. $$1 - \frac{\bar{x} - \frac{k}{N}}{k - \frac{k}{N}}$$

Examples

# NOT RUN {
pbmc_small
# As CCA requires two datasets, we will split our test object into two just for this example
pbmc1 <- SubsetData(pbmc_small, cells.use = [e-mail][1:40])
pbmc2 <- SubsetData(pbmc_small, cells.use = [e-mail][41:80])
[e-mail]$group <- "group1"
[e-mail]$group <- "group2"
pbmc_cca <- RunCCA(pbmc1,pbmc2)
pbmc_cca <- AlignSubspace(pbmc_cca, reduction.type = "cca",
                          grouping.var = "group", dims.align = 1:5)
CalcAlignmentMetric(pbmc_cca, reduction.use = "cca.aligned",
                    dims.use = 1:5, grouping.var =  "group")
# }
# NOT RUN {
# }