Entering edit mode
Hey! I'm doing some analyses with Limma. I have 5 cluster and I want to find genes up and down regulate in the first one versus the others 4, to identify the ones that are typical of this cluster.
mm <- model.matrix(~0 + cluster_number)
y <- voom(d0, mm, plot = T)
tmp <- voom(d0, mm, plot = T)
fit <- lmFit(y, mm)
contr_1_vs_all <- makeContrasts(cluster_number1 - cluster_number2 - cluster_number3
-cluster_number4 -cluster_number5, levels = colnames(coef(fit)))
tmp_1_vs_all <- contrasts.fit(fit, contr_1_vs_all)
tmp_1_vs_all <- eBayes(tmp_1_vs_all)
top.table_1_vs_all <- topTable(tmp_1_vs_all, n = Inf)
top.table_1_vs_all$Gene <- rownames(top.table_1_vs_all)
top.table_1_vs_all <- top.table_1_vs_all[,c("Gene", names(top.table_1_vs_all)[1:6])]
UpGenes_1_vs_all <- subset(top.table_1_vs_all, logFC > 1 & P.Value < 0.05)
DownGenes_1_vs_all <-subset(top.table_1_vs_all, logFC < -1 & P.Value < 0.05)
This are the commands i'm using but when I run it, I have 105 up regulated genes and 21120 down regulated, and i think is strange. What im doing wrong?
I'm not useing DEseq because I have an arry
The contrast is not balanaced +1-1-1-1-1 is not zero. A better choice is cluster_number1-(cluster_number2+cluster_number3+cluster_number4+cluster_number5)/4.
I did what you suggest but now I have 117 down regulate and 21188 up regulated..
What is d0? Is there any normalization applied? Show full code. Is this single-cell data? Why do you filter on P.Value and not adj.P.Val?
Please show your new code. From the result you report it appears that you have not followed SamGGs advice correctly.
Also, you should be judging DE by FDR rather than by logFC and P.Value.
Avoid adding a package tag of a tool you are not using. It emails the maintainer. In this case you added DESeq2 tag but you say you are not using that tool.