How to do DESeq2 for Kallisto

Question

[Kumar]

Hi, I am trying to perform DESeq2 for Kallisto. I ran Kallisto for 36 PE samples and got results files such as abundance.h5. Now I am trying DESeq2 analysis. I am getting everything correct but while making plotPCA it is making the plot of 6 samples. However, I have a total of 36 samples, I tried to find out the issue but not sure why the script taking only 6 samples for the analysis.

Here is the script:

files <- c("/DataAnalysis/270-aligned/abundance.h5", 
 "/DataAnalysis/272-aligned/abundance.h5") ######just displaying 2 samples here, I have total 36 samples 

names(files) <- paste0("sample", 1:length(files))
txi.kallisto <- tximport(files, type = "kallisto", txOut = TRUE)
head(txi.kallisto$counts)

sampleTable <- data.frame(condition = factor(rep(c("mutant", "control"), each = 3)))
rownames(sampleTable) <- colnames(countdata$counts)

dds <- DESeqDataSetFromTximport(countdata, sampleTable, ~condition)
#PCA plot
vsd <- vst(dds)
class(vsd)
head(colData(vsd))
plotPCA(vsd, "condition")

output

> head(colData(vsd))
DataFrame with 6 rows and 1 column
    condition
     <factor>
sample1    mutant
sample2    mutant
sample3    mutant
sample4   control
sample5   control
sample6   control

[ATpoint]

sampleTable <- data.frame(condition = factor(rep(c("mutant", "control"), each = 3)))

You only specify 6 samples. What is CountData? Are all samples imported by tximport?

[Kumar]

I believe so. All data has been imported by tximport. I changed CountData by txi.kallisto variable but it is showing following error. I am following the process of analysis at https://bioconductor.org/packages/release/bioc/vignettes/tximport/inst/doc/tximport.html. Where do I need to specify samples?

> dds <- DESeqDataSetFromTximport(txi.kallisto, sampleTable, ~condition)
Error in DESeqDataSetFromMatrix(countData = counts, colData = colData,  : 
ncol(countData) == nrow(colData) is not TRUE

[swbarnes2]

Well, you are hand making your ColData there with only 6 samples. If you have count values for 36 samples, how can they possibly be the same? Ar you just copy-pasting that code from some tutorial?

[Kumar]

I am not catching entirely. I am following the tutorial I am new in this analysis. I tried to change but not getting improved. If you could help me to improve the code.

sampleTable <- data.frame(condition = 
factor(rep(c("mutant","mutant","mutant","mutant","mutant","mutant","mutant","mutant","mutant",
                                             "mutant","mutant","mutant","mutant","mutant","mutant","mutant","mutant","mutant",
                                               "control","control","control","control","control","control","control","control",
                                               "control","control","control","control","control","control","control","control",
                                              "control","control"), each = 36)))

rownames(sampleTable) <- colnames(txi.kallisto$counts)

dds <- DESeqDataSetFromTximport(txi.kallisto, sampleTable, ~condition)

error:

Error in `.rowNamesDF<-`(x, value = value) : invalid 'row.names' length

[ATpoint]

Remove the rep(c(....),each=36) part. If you explicitely write down each factor level then you don't have to rep it.

[Kumar]

I got it! if I write with rep as following:

sampleTable <- data.frame(condition = factor(rep(c("mutant", "control"), each = 18)))

[Kumar]

However, I need one more help in the following code. What should I use here instead of ColData.

vsd <- vst(dds)
class(vsd)
head(ColData(vsd))
plotPCA(vsd, "condition")

[ATpoint]

Want the values? assay(vsd), read the manual!

[swbarnes2]

SampleTable is 1296 lines long. Are you so confident in your coding skills that you didn't even look at it before trying to use it?

[Kumar]

I did not know about it. If you could really want to help, you could have sent the link instead of commenting on me. I mentioned that I am new about the analysis and do not know about the coding either.