主要涉及两种校正方式(NormlizeData/Sctansform)以及针对大数据集(including >200,000 cells)进行的数据预处理。
数据集整合是基于anchor,而anchor的产生需要对不同数据集进行校正后计算,不同的校正方式,产生不同的anchor。
基于NormalizeData()的校正(首选)
LC.list <- lapply(X = LC.list, FUN = function(x) {
x[["percent.mt"]] <- PercentageFeatureSet(x, pattern = "^mt-")
x <- subset(x, subset = nFeature_RNA > 500 & nFeature_RNA < 5000 & percent.mt < 20)
x <- NormalizeData(x)
x <- FindVariableFeatures(x, selection.method = "vst", nfeatures = 2000)
})
features <- SelectIntegrationFeatures(object.list = LC.list, nfeatures=2000)
LC.anchors <- FindIntegrationAnchors(object.list = LC.list, anchor.features = features, normalization.method = "LogNormalize")
LC.integrated <- IntegrateData(anchorset = LC.anchors, normalization.method = "LogNormalize")
#校正后的降维、聚类
DefaultAssay(LC.integrated) <- "integrated"
LC.integrated <- ScaleData(LC.integrated, verbose = FALSE)
1万+



被折叠的 条评论
为什么被折叠?



