testNhoods: Perform differential neighbourhood abundance testing

View source: R/testNhoods.R

testNhoodsR Documentation

Perform differential neighbourhood abundance testing

Description

This will perform differential neighbourhood abundance testing after cell counting.

Arguments

x

A Milo object with a non-empty nhoodCounts slot.

design

A formula or model.matrix object describing the experimental design for differential abundance testing. The last component of the formula or last column of the model matrix are by default the test variable. This behaviour can be overridden by setting the model.contrasts argument

design.df

A data.frame containing meta-data to which design refers to

min.mean

A scalar used to threshold neighbourhoods on the minimum average cell counts across samples.

model.contrasts

A string vector that defines the contrasts used to perform DA testing.

fdr.weighting

The spatial FDR weighting scheme to use. Choice from max, neighbour-distance, graph-overlap or k-distance (default). If none is passed no spatial FDR correction is performed and returns a vector of NAs.

robust

If robust=TRUE then this is passed to edgeR and limma which use a robust estimation for the global quasilikelihood dispersion distribution. See edgeR and Phipson et al, 2013 for details.

norm.method

A character scalar, either "logMS", "TMM" or "RLE". The "logMS" method normalises the counts across samples using the log columns sums of the count matrix as a model offset. "TMM" uses the trimmed mean of M-values normalisation as described in Robinson & Oshlack, 2010, whilst "RLE" uses the relative log expression method by Anders & Huber, 2010, to compute normalisation factors relative to a reference computed from the geometric mean across samples. The latter methods provides a degree of robustness against false positives when there are very large compositional differences between samples.

reduced.dim

A character scalar referring to the reduced dimensional slot used to compute distances for the spatial FDR. This should be the same as used for graph building.

Details

This function wraps up several steps of differential abundance testing using the edgeR functions. These could be performed separately for users who want to exercise more contol over their DA testing. By default this function sets the lib.sizes to the colSums(x), and uses the Quasi-Likelihood F-test in glmQLFTest for DA testing. FDR correction is performed separately as the default multiple-testing correction is inappropriate for neighbourhoods with overlapping cells.

Value

A data.frame of model results, which contain:

logFC:

Numeric, the log fold change between conditions, or for an ordered/continous variable the per-unit change in (normalized) cell counts per unit-change in experimental variable.

logCPM:

Numeric, the log counts per million (CPM), which equates to the average log normalized cell counts across all samples.

F:

Numeric, the F-test statistic from the quali-likelihood F-test implemented in edgeR.

PValue:

Numeric, the unadjusted p-value from the quasi-likelihood F-test.

FDR:

Numeric, the Benjamini & Hochberg false discovery weight computed from p.adjust.

Nhood:

Numeric, a unique identifier corresponding to the specific graph neighbourhood.

SpatialFDR:

Numeric, the weighted FDR, computed to adjust for spatial graph overlaps between neighbourhoods. For details see graphSpatialFDR.

Author(s)

Mike Morgan

Examples

library(SingleCellExperiment)
ux.1 <- matrix(rpois(12000, 5), ncol=400)
ux.2 <- matrix(rpois(12000, 4), ncol=400)
ux <- rbind(ux.1, ux.2)
vx <- log2(ux + 1)
pca <- prcomp(t(vx))

sce <- SingleCellExperiment(assays=list(counts=ux, logcounts=vx),
                            reducedDims=SimpleList(PCA=pca$x))

milo <- Milo(sce)
milo <- buildGraph(milo, k=20, d=10, transposed=TRUE)
milo <- makeNhoods(milo, k=20, d=10, prop=0.3)
milo <- calcNhoodDistance(milo, d=10)

cond <- rep("A", ncol(milo))
cond.a <- sample(1:ncol(milo), size=floor(ncol(milo)*0.25))
cond.b <- setdiff(1:ncol(milo), cond.a)
cond[cond.b] <- "B"
meta.df <- data.frame(Condition=cond, Replicate=c(rep("R1", 132), rep("R2", 132), rep("R3", 136)))
meta.df$SampID <- paste(meta.df$Condition, meta.df$Replicate, sep="_")
milo <- countCells(milo, meta.data=meta.df, samples="SampID")

test.meta <- data.frame("Condition"=c(rep("A", 3), rep("B", 3)), "Replicate"=rep(c("R1", "R2", "R3"), 2))
test.meta$Sample <- paste(test.meta$Condition, test.meta$Replicate, sep="_")
rownames(test.meta) <- test.meta$Sample
da.res <- testNhoods(milo, design=~Condition, design.df=test.meta[colnames(nhoodCounts(milo)), ], norm.method="TMM")
da.res


MikeDMorgan/miloR documentation built on Aug. 7, 2022, 8:21 a.m.