I have 2 scRNA seq datasets that I'm trying to merge and correct for batch effect by mnnCorrect. I'm generally using Seurat for my analysis but I've used scran/scater for the normalization step. My original data sets do not have the same number of genes but I've subsetted the data based on highly variable genes shared between the two dataset ( 698 genes). I still get the error indicating that the number of rows is not the same across batches. Any idea why? 

Here's my code... 

Cont_MF <- Read10X(data.dir = "/Users/snejat/Documents/Sarah_scData/Epelman_Sarah_cont/mm10_TDTomato")

MI_MF <- Read10X(data.dir = "/Users/snejat/Documents/Sarah_scData/Epelman_Sarah_M1_MFS/mm10_TDTomato")

dim(MI_MF)  # 28001  4697

dim(Cont_MF) #28001  1806

Cont_MFs <- CreateSeuratObject(raw.data = Cont_MF, min.cells = 3, min.genes = 200, 

                            project = "Cont_MFs")

MI_MFs <- CreateSeuratObject(raw.data = MI_MF, min.cells = 3, min.genes = 200, 

                       project = "MI_MFs")

mito.genes <- grep(pattern = "^mt-", x = rownames(x = Cont_MFs@data), value = TRUE)

percent.mito <- Matrix::colSums(Cont_MFs@raw.data[mito.genes, ])/Matrix::colSums(Cont_MFs@raw.data)

mito.genes <- grep(pattern = "^mt-", x = rownames(x = MI_MFs@data), value = TRUE)

percent.mito <- Matrix::colSums(MI_MFs@raw.data[mito.genes, ])/Matrix::colSums(MI_MFs@raw.data)

length(mito.genes) #13

Cont_MFs <- AddMetaData(object = Cont_MFs, metadata = percent.mito, col.name = "percent.mito")

MI_MFs <- AddMetaData(object = MI_MFs, metadata = percent.mito, col.name = "percent.mito")

Cont_MFs2 <- FilterCells(object = Cont_MFs, subset.names = c("nGene", "percent.mito"), 

                     low.thresholds = c(1100, -Inf), high.thresholds = c(5500, 0.1))

MI_MFs2 <- FilterCells(object = MI_MFs, subset.names = c("nGene", "percent.mito"), 

                     low.thresholds = c(1000, -Inf), high.thresholds = c(4800, 0.1))

dim(Cont_MFs2@data) # 14263  1730

dim(MI_MFs2@data) #14784  4626

Cont_MFs2_meta <-Cont_MFs2@meta.data

MI_MFs2_meta <-MI_MFs2@meta.data

Cont_MFs2 <- as.matrix(Cont_MFs2@data)

MI_MFs2 <- as.matrix(MI_MFs2@data)

Cont_sce <- SingleCellExperiment(assays = list(counts = Cont_MFs2))

MI_sce <- SingleCellExperiment(assays = list(counts = MI_MFs2))

#compute sum factors for normalization

Cont_sce <- scran::computeSumFactors(Cont_sce)

MI_sce <- scran::computeSumFactors(MI_sce)

#normalize data - uses the above sum factors

Cont_sce <- normalize(Cont_sce)

MI_sce <- normalize(MI_sce)

Cont_sce_norm <- as.matrix(exprs(Cont_sce))

MI_sce_norm <- as.matrix(exprs(MI_sce))

#reverse the log2 + 1

Cont_norm_nolog <- (2^(Cont_sce_norm)) -1

MI_norm_nolog <- (2^(MI_sce_norm)) -1

#natural log + 1

Cont_norm_ln <- log(Cont_norm_nolog + 1)

MI_norm_ln <- log(MI_norm_nolog + 1)

#load your original data into seurat

Cont_MFs2 <- CreateSeuratObject(raw.data = Cont_MFs2, min.cells = 3, min.genes = 300, project = "Cont_MI", meta.data = Cont_MFs2_meta)

MI_MFs2 <- CreateSeuratObject(raw.data = MI_MFs2, min.cells = 3, min.genes = 300, project = "MI_MI", meta.data = MI_MFs2_meta)

#load your newly normalized (natural log + 1) into the seurat.raw@data slot

Cont_MFs2@data <- Cont_norm_ln

MI_MFs2@data <- MI_norm_ln

Cont_MFs2 <- FindVariableGenes(object = Cont_MFs2, mean.function = ExpMean, dispersion.function = LogVMR, do.plot = TRUE, x.low.cutoff = 0.0125, x.high.cutoff = 3, y.cutoff = 0.5)

length( Cont_MFs2@var.genes)

[1] 1165

MI_MFs2 <- FindVariableGenes(object = MI_MFs2, mean.function = ExpMean, dispersion.function = LogVMR, do.plot = TRUE, x.low.cutoff = 0.0125, x.high.cutoff = 3, y.cutoff = 0.5)

length(MI_MFs2@var.genes)

[1] 1020

common_hvgs <- intersect(Cont_MFs2@var.genes,MI_MFs2@var.genes)

length(common_hvgs) # 698

Cont_MFs2_matrix <- as.matrix(Cont_MFs2@data)

MI_MFs2_matrix <- as.matrix(MI_MFs2@data)

Cont_MFs2_matrix_sample <- as.matrix(Cont_MFs2_matrix [common_hvgs, ])

MI_MFs2_matrix_sample <- as.matrix(MI_MFs2_matrix [common_hvgs, ])

dim(Cont_MFs2_matrix_sample)

[1]  698 1730

dim(MI_MFs2_matrix_sample)

[1]  698 4626

mnn_correction <- mnnCorrect(Cont_MFs2_matrix_sample , MI_MFs2_matrix_sample, k=20, sigma=1, cos.norm.in=FALSE, cos.norm.out=FALSE, subset.row=common_hvgs, var.adj = TRUE )

Error in mnnCorrect(Cont_MFs2_matrix_sample, MI_MFs2_matrix_sample, k = 20,  :
  number of rows is not the same across batches

Here's the session info

> sessionInfo()
R version 3.4.4 (2018-03-15)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS Sierra 10.12.6

Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib

locale:
[1] en_CA.UTF-8/en_CA.UTF-8/en_CA.UTF-8/C/en_CA.UTF-8/en_CA.UTF-8

attached base packages:
[1] parallel  stats4    stats     graphics  grDevices utils     datasets
[8] methods   base

other attached packages:
 [1] BiocInstaller_1.28.0       Rtsne_0.13
 [3] scales_0.5.0               limma_3.34.9
 [5] scater_1.6.3               scran_1.6.9
 [7] SingleCellExperiment_1.0.0 SummarizedExperiment_1.8.1
 [9] DelayedArray_0.4.1         matrixStats_0.53.1
[11] Biobase_2.38.0             GenomicRanges_1.30.3
[13] GenomeInfoDb_1.14.0        IRanges_2.12.0
[15] S4Vectors_0.16.0           BiocGenerics_0.24.0
[17] BiocParallel_1.12.0        dplyr_0.7.6
[19] Seurat_2.3.3               Matrix_1.2-14
[21] cowplot_0.9.2              ggplot2_3.0.0

loaded via a namespace (and not attached):
  [1] shinydashboard_0.7.0   R.utils_2.6.0          tidyselect_0.2.4
  [4] RSQLite_2.1.1          AnnotationDbi_1.40.0   htmlwidgets_1.2
  [7] grid_3.4.4             trimcluster_0.1-2      ranger_0.10.1
 [10] munsell_0.5.0          codetools_0.2-15       ica_1.0-2
 [13] DT_0.4                 statmod_1.4.30         withr_2.1.2
 [16] colorspace_1.3-2       knitr_1.20             rstudioapi_0.7
 [19] geometry_0.3-6         ROCR_1.0-7             robustbase_0.93-1
 [22] dtw_1.20-1             dimRed_0.1.0           lars_1.2
 [25] tximport_1.6.0         GenomeInfoDbData_1.0.0 mnormt_1.5-5
 [28] bit64_0.9-7            rhdf5_2.22.0           ipred_0.9-6
 [31] diptest_0.75-7         R6_2.2.2               ggbeeswarm_0.6.0
 [34] VGAM_1.0-5             locfit_1.5-9.1         flexmix_2.3-14
 [37] DRR_0.0.3              bitops_1.0-6           assertthat_0.2.0
 [40] SDMTools_1.1-221       nnet_7.3-12            beeswarm_0.2.3
 [43] gtable_0.2.0           ddalpha_1.3.4          timeDate_3043.102
 [46] rlang_0.2.1            CVST_0.2-2             scatterplot3d_0.3-41
 [49] RcppRoll_0.3.0         splines_3.4.4          lazyeval_0.2.1
 [52] ModelMetrics_1.1.0     acepack_1.4.1          broom_0.4.5
 [55] checkmate_1.8.5        reshape2_1.4.3         abind_1.4-5
 [58] backports_1.1.2        httpuv_1.4.4.2         Hmisc_4.1-1
 [61] caret_6.0-80           tools_3.4.4            lava_1.6.2
 [64] psych_1.8.4            gplots_3.0.1           RColorBrewer_1.1-2
 [67] proxy_0.4-22           dynamicTreeCut_1.63-1  ggridges_0.5.0
 [70] Rcpp_0.12.17           plyr_1.8.4             base64enc_0.1-3
 [73] progress_1.2.0         zlibbioc_1.24.0        purrr_0.2.5
 [76] RCurl_1.95-4.10        prettyunits_1.0.2      rpart_4.1-13
 [79] pbapply_1.3-4          viridis_0.5.1          zoo_1.8-2
 [82] sfsmisc_1.1-2          cluster_2.0.7-1        magrittr_1.5
 [85] data.table_1.11.4      lmtest_0.9-36          RANN_2.5.1
 [88] mvtnorm_1.0-8          fitdistrplus_1.0-9     mime_0.5
 [91] xtable_1.8-2           XML_3.98-1.11          mclust_5.4.1
 [94] gridExtra_2.3          compiler_3.4.4         biomaRt_2.34.2
 [97] tibble_1.4.2           KernSmooth_2.23-15     R.oo_1.22.0
[100] htmltools_0.3.6        segmented_0.5-3.0      Formula_1.2-3
[103] snow_0.4-2             tidyr_0.8.1            tclust_1.4-1
[106] lubridate_1.7.4        DBI_1.0.0              diffusionMap_1.1-0
[109] magic_1.5-8            MASS_7.3-50            fpc_2.1-11
[112] R.methodsS3_1.7.1      gdata_2.18.0           metap_0.9
[115] bindr_0.1.1            gower_0.1.2            igraph_1.2.1
[118] pkgconfig_2.0.1        sn_1.5-2               numDeriv_2016.8-1
[121] foreign_0.8-70         recipes_0.1.3          foreach_1.4.4
[124] vipor_0.4.5            XVector_0.18.0         prodlim_2018.04.18
[127] stringr_1.3.1          digest_0.6.15          tsne_0.1-3
[130] htmlTable_1.12         edgeR_3.20.9           kernlab_0.9-26
[133] shiny_1.1.0            gtools_3.8.1           modeltools_0.2-21
[136] rjson_0.2.20           nlme_3.1-137           bindrcpp_0.2.2
[139] viridisLite_0.3.0      pillar_1.2.3           lattice_0.20-35
[142] DEoptimR_1.0-8         survival_2.42-4        glue_1.2.0
[145] FNN_1.1                png_0.1-7              prabclus_2.2-6
[148] iterators_1.0.9        bit_1.1-14             class_7.3-14
[151] stringi_1.2.3          mixtools_1.1.0         blob_1.1.1
[154] doSNOW_1.0.16          latticeExtra_0.6-28    caTools_1.17.1
[157] memoise_1.1.0          irlba_2.3.2            ape_5.1

There are several points here.

Firstly, you should be using the latest version of scran (1.8.3) and, more generally, of Bioconductor. The error you talk about may have already been fixed in the latest version. Even if there is still a bug in the latest version, you will have to upgrade anyway to get the bugfix, so you might as well do it now.

If the bug persists upon upgrading the packages; the error that you've described should be impossible based on an inspection of the code. You'll have to make a minimum working example that reproduces the error (e.g., based on some simulated data) in order for me to investigate it.

Secondly, if you're going to upgrade, you might want to consider using the development version of scran:

https://www.bioconductor.org/developers/how-to/useDevel/

... which contains a much-improved fastMNN function. This is a faster and more robust version of the MNN correction algorithm, see https://github.com/MarioniLab/FurtherMNN2018 for theoretical details and performance comparisons. Also see http://bioconductor.org/packages/devel/workflows/vignettes/simpleSingleCell/inst/doc/work-5-mnn.html for how to use it.

Thirdly; I can't help but feel you're making life more difficult for yourself by manually converting between SingleCellExperiment and Seurat objects. There are a plethora of HVG-calling functions in scran - trendVar and decomposeVar, technicalCV2, improvedCV2, DM. If you want negative binomial dispersions, you should also be able to use SingleCellExperiment objects in DESeq2, or use convertTo to obtain a DGEList for use in edgeR, and then use the residuals of the mean-dispersion trend for identifying highly variable genes. I'd be surprised if none of these strategies were able to serve your needs.