Negative average expression value on Dotplot

[trueth1206]

Hi, Thank you for creating this excellent tool for single cell RNA sequencing analysis. I do not quite understand why the average expression value on my dotplot starts from -1. Could anybody help me?

[timoast]

Are you plotting integrated data? There can be negative values in the integrated data and this is expected, and has been stated in other closed issues. In general we recommend differential expression and visualization of the uncorrected data.

[trueth1206]

Hi, Thank you for your reply. Apologize for asking a similar question. Yes, I used an integrated data. But I used the RNA assay for dot plotting. Will I still get a negative value when using the RNA assay in integrated data?

[timoast]

If you use the RNA assay containing uncorrected data there should not be negative average expression values. Can you post the full code you're running?

[trueth1206]

Hi, As you mentioned about it, I went back to check my data before integration. I found that the negative value exited before the integration of the two conditions. Here is the code I am using to manage a single condition of 4 biological replicates. It ended up with the plot below:

library("Seurat")
library("dplyr")
library("mclust")
library("cowplot")

D0_1.dgecounts <-
 readRDS("~/Desktop/snRNAseq_analysis/zUMIs/D0_1/expression/D0_1.dgecounts.rds")
D0_2.dgecounts <-
 readRDS("~/Desktop/snRNAseq_analysis/zUMIs/D0_2/expression/D0_2.dgecounts.rds")
D0_3.dgecounts <- 
readRDS("~/Desktop/snRNAseq_analysis/zUMIs/D0_3/expression/D0_3.dgecounts.rds")
D0_4.dgecounts <- 
readRDS("~/Desktop/snRNAseq_analysis/zUMIs/D0_4/expression/D0_4.dgecounts.rds")

rawCountsD0_1 <- D0_1.dgecounts[["umicount"]][["inex"]][["all"]]
seuratRawObjD0_1 <- CreateSeuratObject(rawCountsD0_1, min.cells = 3, min.features = 400)
seuratRawObjD0_1

rawCountsD0_2 <- D0_2.dgecounts[["umicount"]][["inex"]][["all"]]
seuratRawObjD0_2 <- CreateSeuratObject(rawCountsD0_2, min.cells = 3, min.features = 400)
seuratRawObjD0_2

rawCountsD0_3 <- D0_3.dgecounts[["umicount"]][["inex"]][["all"]]
seuratRawObjD0_3 <- CreateSeuratObject(rawCountsD0_3, min.cells = 3, min.features = 400)
seuratRawObjD0_3

rawCountsD0_4 <- D0_4.dgecounts[["umicount"]][["inex"]][["all"]]
seuratRawObjD0_4 <- CreateSeuratObject(rawCountsD0_4, min.cells = 3, min.features = 400)
seuratRawObjD0_4

seuratRawObjD0_1$D0 <- "D0_1"
seuratRawObjD0_2$D0 <- "D0_2"
seuratRawObjD0_3$D0 <- "D0_3"
seuratRawObjD0_4$D0 <- "D0_4"

D0_M1 <- 
merge(seuratRawObjD0_1, seuratRawObjD0_2, add.cell.ids = c("1", "2"), 
project = "SeuratProject")

D0_M2 <- 
merge(seuratRawObjD0_3, seuratRawObjD0_4, add.cell.ids = c("3", "4"), 
project = "SeuratProject")

seuratRawObj <- merge(D0_M1, D0_M2, project = "SeuratProject")
seuratRawObj

seuratRawObj[["percent.mt"]] <- PercentageFeatureSet(seuratRawObj, features = 
c("ENSMUSG00000064336", "ENSMUSG00000064337", "ENSMUSG00000064339", 
"ENSMUSG00000064341", "ENSMUSG00000064343", "ENSMUSG00000064344", 
"ENSMUSG00000064345", "ENSMUSG00000064347", 
"ENSMUSG00000064348","ENSMUSG00000064349", "ENSMUSG00000064351", 
"ENSMUSG00000064356", "ENSMUSG00000064360", "ENSMUSG00000065947", 
"ENSMUSG00000064365", "ENSMUSG00000064367", "ENSMUSG00000064368", 
"ENSMUSG00000064369", "ENSMUSG00000064370", "ENSMUSG00000064371", 
"ENSMUSG00000064372"))

seuratFilteredObj <- 
subset(seuratRawObj, subset = nFeature_RNA > 400 & nFeature_RNA < 7500 & percent.mt < 10)
seuratFilteredObj

seuratNormalisedObj <- NormalizeData(object = seuratFilteredObj, 
normalization.method = "LogNormalize", scale.factor = 10000)

seuratNormalised2Obj <- FindVariableFeatures(seuratNormalisedObj, 
selection.method = "vst", nfeatures = 2000)
top10 <- head(VariableFeatures(seuratNormalised2Obj), 10)
plot3 <- VariableFeaturePlot(seuratNormalised2Obj)
plot4 <- LabelPoints(plot = plot3, points = top10, repel = TRUE)
CombinePlots(plots = list(plot3, plot4))

all.genes <- rownames(seuratNormalised2Obj)
seuratScaledObj <- 
ScaleData(object = seuratNormalised2Obj, features = all.genes)

seuratScaled2Obj <- RunPCA(object = seuratScaledObj, 
features = VariableFeatures(object = seuratScaledObj), do.print = TRUE, 
pcs.print = 1:5, genes.print = 5)

print(seuratScaled2Obj[["pca"]], dims = 1:5, nfeatures = 5)

VizDimLoadings(seuratScaled2Obj, dims = 1:2, reduction = "pca")

DimPlot(seuratScaled2Obj, reduction = "pca")

seuratJackStraw <- JackStraw(seuratScaled2Obj, num.replicate = 100)
scoreJackStraw <- ScoreJackStraw(seuratJackStraw, dims = 1:20)

ElbowPlot(scoreJackStraw, ndims = 30)

seuratFindNeighbors <- FindNeighbors(
  scoreJackStraw,
  reduction = "pca",
  dims = 1:20,
  do.plot = FALSE,
  graph.name = NULL,
  k.param = 30
)
seuratFindClusters <- FindClusters(object = seuratFindNeighbors)

seusetTSNE <- RunTSNE(seuratFindClusters, reduction = "pca", 
dims = 1:20, tsne.method = "Rtsne")

TSNEPlot(object = seusetTSNE, label.size = 4, pt.size = 0.5, label = TRUE)

DefaultAssay(seuratFindClusters) <- "RNA"

markers.to.plot <- c("ENSMUSG00000030781", "ENSMUSG00000018459", 
"ENSMUSG00000024650", "ENSMUSG00000021490", "ENSMUSG00000031441", 
"ENSMUSG00000041052", "ENSMUSG00000040405", "ENSMUSG00000027962", 
"ENSMUSG00000020914", "ENSMUSG00000031004", "ENSMUSG00000004655", 
"ENSMUSG00000029082", "ENSMUSG00000028289", "ENSMUSG00000027202", 
"ENSMUSG00000030963", "ENSMUSG00000031766", "ENSMUSG00000054640", 
"ENSMUSG00000023013", "ENSMUSG00000031891", "ENSMUSG00000104445", 
"ENSMUSG00000028238", "ENSMUSG00000006574", "ENSMUSG00000020651", 
"ENSMUSG00000029648", "ENSMUSG00000062960", "ENSMUSG00000026365", 
"ENSMUSG00000070645", "ENSMUSG00000026395", "ENSMUSG00000006649")

DotPlot(seuratFindClusters, features = rev(markers.to.plot), cols = c("blue", "red"), 
dot.scale = 8) + RotatedAxis()

Thank you again for your kindly help.

[timoast]

I found that the negative value exited before the integration of the two conditions

If you have negative values in your raw data I think something odd has happened in your upstream processing

In any case, if the negative values are present in the data then this is not an issue with the DotPlot function and so I will close this now, I suggest you take a look at how the data is being processed upstream to result in negative values

[Tushar-87]

could it be the problem of slot in the RNA assay? For a dot plot we just need the normalized data with no additional scaling...