Entering edit mode
Hello,
For running the DESeq2 I am using the below command to do the t.test for determining the p-value, to separate those genes which are "significant" expressed, I also used the Excell sheet for determining the "Significant" genes expression, but in both cases, some of the genes which are significantly expressed in the original data, are expressing "Non-significantly" in my analysis. is there someone to advise me where would be the problems? Many thanks
Pvalue=c()
carrierindex=which(CountsNorm$condition=="carrier")
noncarrierindex=which(CountsNorm$condition=="noncarrier")
for (i in 1:nrow(Counts)){
Pvalue=c(Pvalue,t.test(Counts[i,carrierindex],Counts[i,noncarrierindex])$p.value)
}
That command has nothing to do with DESeq.
Applying a simple t-test to RNAseq counts is wrong.
so could you please advise me what kinds of test should I apply?
Can you explain why you won't use DESeq2? Why are you tagging your questions with it and not using it?
I used the below commands for running the DESEq2,
library(DESeq2)
count_matrix <- as.matrix(read.csv("~/Desktop/sample.csv", row.names = 1))
head(count_matrix, 2)
View(count_matrix)
condition <- factor(c(rep("carrier",3), rep("noncarrier", 3)))
sampleTable <- data.frame(condition = as.factor(condition))
rownames(sampleTable) <- colnames(count_matrix)
sampleTable
deseq <- DESeqDataSetFromMatrix(countData = count_matrix,
deseq$condition <- relevel(dds$condition, ref = "noncarrier")
d.deseq <- DESeq(deseq)
res<- results(d.deseq)
res<- results(d.deseq, alpha = 0.05 )
summary(res)
but the results are as below, and the DEGs are very less:
out of 32159 with nonzero total read count
adjusted p-value < 0.05
LFC > 0 (up) : 23, 0.072%
LFC < 0 (down) : 10, 0.031%
outliers [1] : 168, 0.52%
low counts [2] : 2542, 7.9%
(mean count < 0)
[1] see 'cooksCutoff' argument of ?results
[2] see 'independentFiltering' argument of ?results
could you please help me to know where my mistake is, it is about 6 months since I am trying different ways to solve the problems, that is why I thought maybe by using t.test I would be able to solve the problem, I would really appreciate your advise and help in my this tough time.
This is likely not a DESeq2 problem. Either your study is underpowered (3 vs 3 definitely is) or batch effects make things difficult. The vignette covers diagnosis of batch effects via PCA. In any case, switching to underpowered tests like t or Wilcox-test is not going to provide any benefit.
Many thanks for your answer. sorry, I did not get what it means by underpowered or batch effects, could you please make it clear? Thanks in advance
In this case I recommend taking courses or tutorials on basics of RNA-seq and stats/experimental design so you get a background before going into analysis. Among other resources, HBC has great tutorials: https://hbctraining.github.io/DGE_workshop/lessons/04_DGE_DESeq2_analysis.html
You can Google those terms. You can't expect a discussion board to do everything for you.