Correcting batch effects in single-cell RNA-seq data

opts_chunk$set(error=FALSE, message=FALSE, warning=FALSE)


Batch effects refer to differences between data sets generated at different times or in different laboratories. These often occur due to uncontrolled variability in experimental factors, e.g., reagent quality, operator skill, atmospheric ozone levels. The presence of batch effects can interfere with downstream analyses if they are not explicitly modelled. For example, differential expression analyses typically use a blocking factor to absorb any batch-to-batch differences.

For single-cell RNA sequencing (scRNA-seq) data analyses, explicit modelling of the batch effect is less relevant. Manny common downstream procedures for exploratory data analysis are not model-based, including clustering and visualization. It is more generally useful to have methods that can remove batch effects to create an corrected expression matrix for further analysis. This follows the same strategy as, e.g., the removeBatchEffect() function in the r Biocpkg("limma") package [@ritchie2015limma].

Batch correction methods designed for bulk genomics data usually require knowledge of the other factors of variation. This is usually not known in scRNA-seq experiments where the aim is to explore unknown heterogeneity in cell populations. The r Biocpkg("batchelor") package implements batch correction methods that do not rely on a priori knowledge about the population structure.

Setting up demonstration data

To demonstrate, we will use two brain data sets [@tasic2016adult;@zeisel2015brain] from the r Biocpkg("scRNAseq") package. A more thorough explanation of each of these steps is available in the book.

sce1 <- ZeiselBrainData()
sce2 <- TasicBrainData()

We apply some quick-and-dirty quality control to both datasets, using the outlier detection strategy from the r Biocpkg('scuttle') package.

sce1 <- addPerCellQC(sce1, subsets=list(Mito=grep("mt-", rownames(sce1))))
qc1 <- quickPerCellQC(colData(sce1), sub.fields="subsets_Mito_percent")
sce1 <- sce1[,!qc1$discard]

sce2 <- addPerCellQC(sce2, subsets=list(Mito=grep("mt_", rownames(sce2))))
qc2 <- quickPerCellQC(colData(sce2), sub.fields="subsets_Mito_percent")
sce2 <- sce2[,!qc2$discard]

Some preprocessing is required to render these two datasets comparable. We subset to the common subset of genes:

universe <- intersect(rownames(sce1), rownames(sce2))
sce1 <- sce1[universe,]
sce2 <- sce2[universe,]

We compute log-normalized expression values using library size-derived size factors for simplicity. (More complex size factor calculation methods are available in the r Biocpkg("scran") package.)

out <- multiBatchNorm(sce1, sce2)
sce1 <- out[[1]]
sce2 <- out[[2]]

Finally, we identify the top 5000 genes with the largest biological components of their variance. We will use these for all high-dimensional procedures such as PCA and nearest-neighbor searching.

dec1 <- modelGeneVar(sce1)
dec2 <- modelGeneVar(sce2)
combined.dec <- combineVar(dec1, dec2)
chosen.hvgs <- getTopHVGs(combined.dec, n=5000)

As a diagnostic, we check that there actually is a batch effect across these datasets by checking that they cluster separately. Here, we combine the two SingleCellExperiment objects without any correction using the NoCorrectParam() flag, and we informally verify that cells from different batches are separated using a $t$-SNE plot.

combined <- correctExperiments(A=sce1, B=sce2, PARAM=NoCorrectParam())

combined <- runPCA(combined, subset_row=chosen.hvgs)
combined <- runTSNE(combined, dimred="PCA")
plotTSNE(combined, colour_by="batch")

Function organization

The batch correction functions in r Biocpkg("batchelor") are organized into three levels:

For brevity, we will demonstrate the lowest-level functions in this vignette; however, it is easy to perform the exact same correction with correctExperiments() by switching the above call to, e.g., PARAM=FastMnnParam().

Mutual nearest neighbors


Mutual nearest neighbors (MNNs) are defined as pairs of cells - one from each batch - that are within each other's set of k nearest neighbors. The idea is that MNN pairs across batches refer to the same cell type, assuming that the batch effect is orthogonal to the biological subspace [@haghverdi2018batch]. Once MNN pairs are identified, the difference between the paired cells is used to infer the magnitude and direction of the batch effect. It is then straightforward to remove the batch effect and obtain a set of corrected expression values.

The new, fast method

The fastMNN() function performs a principal components analysis (PCA) on the HVGs to obtain a low-dimensional representation of the input data. MNN identification and correction is performed in this low-dimensional space, which offers some advantages with respect to speed and denoising. This returns a SingleCellExperiment containing a matrix of corrected PC scores is returned, which can be used directly for downstream analyses such as clustering and visualization. (We set the seed as we default to a stochastic method for PCA - see comments on multiBatchPCA() below.)

f.out <- fastMNN(A=sce1, B=sce2, subset.row=chosen.hvgs)
str(reducedDim(f.out, "corrected"))

The batch of origin for each row/cell in the output matrix is also returned:


Another way to call fastMNN() is to specify the batch manually. This will return a corrected matrix with the cells in the same order as that in the input combined object. In this case, it doesn't matter as the two batches were concatenated to created combined anyway, but these semantics may be useful when cells from the same batch are not contiguous.

f.out2 <- fastMNN(combined, batch=combined$batch, subset.row=chosen.hvgs)
str(reducedDim(f.out2, "corrected"))

As we can see, cells from the batches are more intermingled in the $t$-SNE plot below. This suggests that the batch effect was removed - assuming, of course, that the intermingled cells represent the same population.

f.out <- runTSNE(f.out, dimred="corrected")
plotTSNE(f.out, colour_by="batch")

We can also obtain per-gene corrected expression values by using the rotation vectors stored in the output. This reverses the original projection used to obtain the initial low-dimensional representation. There are, however, several caveats to using these values for downstream analyses.

cor.exp <- assay(f.out)[1,]
hist(cor.exp, xlab="Corrected expression for gene 1", col="grey80") 

While the default arguments are usually satisfactory, there are many options for running fastMNN(), e.g., to improve speed or to achieve a particular merge order. Refer to the documentation at ?fastMNN or the book for more details.

The old, classic method

The original method described by @haghverdi2018batch is implemented in the mnnCorrect() method. This performs the MNN identification and correction in the gene expression space, and uses a different strategy to overcome high-dimensional noise. mnnCorrect() is called with the same semantics as fastMNN():

# Using fewer genes as it is much slower. 
fewer.hvgs <- head(order(combined.dec$bio, decreasing=TRUE), 500)
classic.out <- mnnCorrect(sce1, sce2, subset.row=fewer.hvgs)

... but returns the corrected gene expression matrix directly, rather than using a low-dimensional representation^[Again, those readers wanting to use the corrected values for per-gene analyses should consider the caveats mentioned previously.]. This is wrapped in a SingleCellExperiment object to store various batch-related metadata.


For scRNA-seq data, fastMNN() tends to be both faster and better at achieving a satisfactory merge. mnnCorrect() is mainly provided here for posterity's sake, though it is more robust than fastMNN() to certain violations of the orthogonality assumptions.

The cluster-based method

In some scenarios, we have already separately characterized the heterogeneity in each batch, identifying clusters and annotating them with relevant biological states. The clusterMNN() function allows us to examine the relationships between clusters across batches, as demonstrated below with the Tasic and Zeisel datasets using the author-provided labels.

# Removing the 'unclassified' cluster, which makes no sense:
not.unclass <- sce2$broad_type!="Unclassified"
clust.out <- clusterMNN(sce1, sce2[,not.unclass],
    clusters=list(sce1$level1class, sce2$broad_type[not.unclass])) 

The most relevant output is extracted from the metadata and describes the "meta-clusters", i.e., groupings of matching clusters corresponding to the same biological state across different batches. <- metadata(clust.out)$cluster

The output object itself is a SingleCellExperiment that can be used for downstream processing in the same manner as the output of fastMNN(). Compared to fastMNN(), clusterMNN() is more faithful with respect to preseving the separation between clusters; however, it will not attempt to adjust for differences in intra-cluster structure between batches. This can generally be considered a more conservative strategy than fastMNN() for merging batches.

Batch rescaling

rescaleBatches() effectively centers the batches in log-expression space on a per-gene basis. This is conceptually equivalent to running removeBatchEffect() with no covariates other than the batch. However, rescaleBatches() achieves this rescaling by reversing the log-transformation, downscaling the counts so that the average of each batch is equal to the smallest value, and then re-transforming. This preserves sparsity by ensuring that zeroes remain so after correction, and mitigates differences in the variance when dealing with counts of varying size between batches^[Done by downscaling, which increases the shrinkage from the added pseudo-count.].

Calling rescaleBatches() returns a corrected matrix of per-gene log-expression values, wrapped in a SummarizedExperiment containin batch-related metadata. This function operates on a per-gene basis so there is no need to perform subsetting (other than to improve speed).

rescale.out <- rescaleBatches(sce1, sce2)

While this method is fast and simple, it makes the strong assumption that the population composition of each batch is the same. This is usually not the case for scRNA-seq experiments in real systems that exhibit biological variation. Thus, rescaleBatches() is best suited for merging technical replicates of the same sample, e.g., that have been sequenced separately.

rescale.out <- runPCA(rescale.out, subset_row=chosen.hvgs,
plotPCA(rescale.out, colour_by="batch")

Alternatively, a more direct linear regression of the batch effect can be performed with regressBatches(). This does not preserve sparsity but uses a different set of tricks to avoid explicitly creating a dense matrix, specifically by using the ResidualMatrix class from the r Biocpkg("ResidualMatrix") package.

regress.out <- regressBatches(sce1, sce2)

Using data subsets

Selecting genes

As shown above, the subset.row= argument will only perform the correction on a subset of genes in the data set. This is useful for focusing on highly variable or marker genes during high-dimensional procedures like PCA or neighbor searches, mitigating noise from irrelevant genes. For per-gene methods, this argument provides a convenient alternative to subsetting the input.

For some functions, it is also possible to set correct.all=TRUE when subset.row= is specified. This will compute corrected values for the unselected genes as well, which is possible once the per-cell statistics are obtained with the gene subset. With this setting, we can guarantee that the output contains all the genes provided in the input.

Restricted correction

Many functions support the restrict= argument whereby the correction is determined using only a restricted subset of cells in each batch. The effect of the correction is then - for want of a better word - "extrapolated" to all other cells in that batch. This is useful for experimental designs where a control set of cells from the same source population were run on different batches. Any difference in the controls between batches must be artificial in origin, allowing us to estimate and remove the batch effect without making further biological assumptions.

# Pretend the first X cells in each batch are controls.
restrict <- list(1:100, 1:200) 
rescale.out <- rescaleBatches(sce1, sce2, restrict=restrict)

Other utilities

Multi-batch normalization

Differences in sequencing depth between batches are an obvious cause for batch-to-batch differences. These can be removed by multiBatchNorm(), which downscales all batches to match the coverage of the least-sequenced batch. This function returns a list of SingleCellExperiment objects with log-transformed normalized expression values that can be directly used for further correction.

normed <- multiBatchNorm(A=sce1, B=sce2,

Downscaling mitigates differences in variance between batches due to the mean-variance relationship of count data. It is achieved using a median-based estimator to avoid problems with composition biases between batches [@lun2016pooling]. Note that this assumes that most genes are not DE between batches, which we try to avoid violating by performing the rescaling on all genes rather than just the HVGs.

Multi-batch PCA

Users can perform a PCA across multiple batches using the multiBatchPCA() function. The output of this function is roughly equivalent to cbinding all batches together and performing PCA on the merged matrix. The main difference is that each sample is forced to contribute equally to the identification of the rotation vectors. This allows small batches with unique subpopulations to participate in the definition of the low-dimensional space.


# Using the same BSPARAM argument as fastMNN(), for speed.
pca.out <- multiBatchPCA(A=sce1, B=sce2, subset.row=chosen.hvgs,

This function is used internally in fastMNN() but can be explicitly called by the user. Reduced dimensions can be input into reducedMNN(), which is occasionally convenient as it allows different correction parameters to be repeated without having to repeat the time-consuming PCA step. We use the IRLBA algorithm by default, which means that we need to set seed to guarantee reproducible results. We also set deferred=TRUE to speed up the calculations when using a custom weighting scheme for the batches - see ?"BiocSingular-options" for more details.

Session information {-}



Try the batchelor package in your browser

Any scripts or data that you put into this service are public.

batchelor documentation built on April 17, 2021, 6:02 p.m.