crazyhottommy
Should I do ScaleData on my integrated data as an input for scclusteval
Dear @crazyhottommy In the preprocess.R script should I add ScaleData before RunPCA as I haven't run it on my integrated data. Or I need to run it on my integrated data? Thanks, Paria
SCTransform replaces NormalizeData, ScaleData and FindVariableFeatures. what normalization did you use?
Thanks for your response. I did NormalizeData, ScaleData before integration. So, there is no need to do it in preprocess.R right?what about adding ScaleData in subsample.R considering that I didn't do SCTransform. It is how I changed the subsample.R meta.data.colnames <- [email protected] %>% colnames() vars.to.regress <- c("percent.mt", "nCount_RNA") vars.to.regress <- vars.to.regress[vars.to.regress %in% meta.data.colnames] object <- ScaleData(object, vars.to.regress = vars.to.regress, model.use = "poisson", use.umi = TRUE) object <- RunPCA(object = object, features = VariableFeatures(object = object), npcs = num.pc) pc.use.meta <- rep(pc.use, length(colnames(object))) names(pc.use.meta) <- colnames(object) object <- AddMetaData(object = object, metadata = pc.use.meta, col.name = "pc.use") object <- RunHarmony(object, reduction.save = "harmony", group.by.vars = c("batchlib", "individual")) object <- UpdateSeuratObject(object) object <- FindNeighbors(object, reduction = "harmony", dims = 1:pc.use, k.param = k.param, nn.eps = nn.eps, verbose = FALSE, force.recalc = TRUE) object <- FindClusters(object = object, n.start = n.start, resolution = resolution, verbose = FALSE, algorithm = 2) return(object) }