How can I normalize miRNA read count from TCGA

Question

[Fluke]

Hi everyone.

I downloaded miRNA expression (Read_Count) from TCGA and then I want to normalized these read count using DESeq2. I have ever used this function to normalized count of RNA-seq from TCGA and it' work.

dds <- DESeqDataSetFromMatrix(countData = round(data), colData = coldData, design = ~TYPE)
sizeFactors(dds)
normalized_counts <- counts(dds, normalized=TRUE)

Now, I use the same code to normalize read count of miRNA expression but I got an error

dds <- DESeqDataSetFromMatrix(countData = round(data), colData = coldData, design = ~TYPE)
sizeFactors(dds)

NULL

How can I solve this situation or are there any method to normalize miRNA expression read count?

[Ram]

How did you get the data object?

[Fluke]

I used TCGAbiolink to download read count data from TCGA instead of read per million (RPM) and then I excluded some sample using my inclusion criteria (I did it in excel) and imported to Rstudio again and created count matrix from the following code

x <- read.csv(" tcga_luad_mirna_data.csv")
x <- x[!duplicated(x$miRNA_ID), ]
x <-  x[!duplicated(x), ]
x2 <- x[,-1]
rownames(x2) <- x[,1]
data <- x2
head(data)
traits <- read.csv("type-MIRNA.csv")
phenodata1 <- traits[,-1]
rownames(phenodata1) <- traits[,1]
head(phenodata1)
coldData <- phenodata1
dds <- DESeqDataSetFromMatrix(countData = round(data), colData = coldData, design = ~TYPE)
sizeFactors(dds)

[Ram]

A few tips:

1. Avoid Excel. It mangles some gene names and makes reproducibility difficult. As far as processing 2D data goes, if you can do it in Excel, you can do it in R.
2. It would be even more helpful to include the code you used to download and pre-process the data - not necessary now that ATPoint has answered your question, but doesn't hurt to have in the future.

[Fluke]

I will follow your suggestions. thanks for your suggestions

Answer 1

[ATpoint]

The size factors are always NULL after construction. You have to run estimateSizeFactors first, see vignette.

[Fluke]

OMG, that's work. Thank a lot!!! and sorry for another stupid question. I am hesitated about what s better for limma between read count and read per million