Dear All,

I have a list with 9 single cell experiment (SCE) objects which I want to integrate using batchelor::fastMNN. The experiment design is: 3 donors gave their blood in 3 different times and their blood was stimulated with a recombinant protein. I have done the data QC and scTransform:vst for normalization, transformation and feature selection. Now I would like to integrate the 9 SCE objects in a single SCE using batchelor::fastMNN. In the function vignette it states: "batch: A vector or factor specifying the batch of origin for all cells when only a single object is supplied in .... This is ignored if multiple objects are present.". This means that if I use as input the list with the 9 SCE, the batch will be ignored and each SCE within the list will be treated as a batch. Now my question: Since I have paired samples, wouldn`t make more sense to merge the 9 SCE before correction (merge_sce), and then within the batchelor::fastMNN do:

# MNN correction
set.seed(123)
mnn_out <- batchelor::fastMNN(merged_sce,
                              subset.row = hvgs_ens,
                              **batch = merged_sce$donor**,
                              assay.type = "vst",
                              auto.merge = T,
                              weights = T, 
                              get.variance = T, d = 50, k = 20, 
                              BSPARAM = BiocSingular::IrlbaParam(deferred = TRUE))

What would be the best way to merge my data?

Thank you for any help. José

If I wanted to first merge paired samples, I would probably combine the 9 SCEs into a single SCE and then use the merge.order argument See ?batchelor::fastMNN for a description of merge.order, in particular how it can be specified as "a list of lists representing a tree structure specifying a hierarchical merge order".