I wanted to run SC3 on normalized (scran/scater) logcounts that were then sent through limma's removeBatchEffect using the follwing script:

library(scater)
library(scran)
library(SC3)

txi.bbsplit <- readRDS("txi-bbsplit.rds")

counts.annotated <- txi.bbsplit$counts

### Removes genes that have no expression at all
counts.annotated <- counts.annotated[rowSums(counts.annotated)>0,]

### Load phenotypes
labels <- read.csv(file = "../118-phenotype-table.csv")

### Leave out the cells that were not added to the pool
labels <- labels[labels$CellNumber %in% as.character(colnames(counts.annotated)),]
labels$X <- NULL

## removed sex as it is complitely contained in the animal number
labels$Sex <- NULL

## removed Date as it is also completely contained in the animal
labels$Date <- NULL

rownames(labels) <- labels$CellNumber
labels$Animal <- factor(labels$Animal)
labels$AmpBatch <- factor(labels$AmpBatch)

pd <- new("AnnotatedDataFrame", data = labels)
sce <- newSCESet(countData = data.matrix(counts.annotated), phenoData = pd)

is.mito <- grepl("^mt.", rownames(sce))


sce <- calculateQCMetrics(sce, feature_controls = list(Mt = is.mito))

libsize.drop <- isOutlier(sce$total_counts, nmads=3, type="lower", log=TRUE)
feature.drop <- isOutlier(sce$total_features, nmads=3, type="lower", log=TRUE)
mito.drop <- isOutlier(sce$pct_counts_Mt, nmads=3, type="higher")
sce <- sce[,!(libsize.drop | feature.drop | mito.drop)]
filtered.out <- data.frame(ByLibSize=sum(libsize.drop), ByFeature=sum(feature.drop),
                           ByMito=sum(mito.drop), Remaining=ncol(sce))

## ave.counts <- calcAverage(sce)

## keep <- ave.counts >= 1

## sce <- sce[keep,]

sce <- computeSumFactors(sce)
sce <- normalize(sce)

library(limma)
logcounts.sce <- logcounts(sce)
logcounts(sce) <- removeBatchEffect(logcounts.sce, batch=sce$AmpBatch)

sce <- sc3_prepare(sce, ks = 2:6)

I get the following error:

Error in sc3_prepare(sce, ks = 2:6) :
  All genes were removed after the gene filter! Stopping now...

The above script runs fine without the batch effect removal step. Any idea why this happens?

Hi, what does removeBatchEffect return? If it remove zeros from your matrix this is not good for SC3 gene filter, which removes genes that have either very high or very low number of zeros. So, if there are no zeros in your data after batch correction, please switch the gene filter off by setting gene_filter = FALSE inside of sc3_prepare.