NA values of count matrix in class DESeqDataSet

Question

[maleknias]

Dear all

Hi

I downloaded a data set in class "RangedSummarizedExperiment" from "https://jhubiostatistics.shinyapps.io/recount/". I want to find differential expression genes. My code is :

load("~/Downloads/rse_gene.Rdata")

class(rse_gene)

**[1] "RangedSummarizedExperiment"

attr(,"package")

[1] "SummarizedExperiment"**

data=colData(rse_gene)

names= names(colData(rse_gene))

write.table(data,file="colData.csv", col.names=names,sep="\t",row.names=FALSE)

data1=fread("~/Downloads/colData.txt")

colData(rse_gene) =DataFrame(data1)

colData(rse_gene)$disease.status = as.factor(colData(rse_gene)$disease.status)

dds <- DESeqDataSet(rse_gene, design = ~ disease.status)

**converting counts to integer mode Error in validObject(.Object) :

invalid class “DESeqDataSet” object: NA values are not allowed in the count matrix In addition: Warning message: In mde(x) : NAs introduced by coercion to integer range**

I use two solution for this problem but both of them were useless:

1- Keep only rows with non-zero counts:

rse_gene <- rse_gene[rowSums(assay(rse_gene)) != 0, ]

2- Replace the NA value by -9 :

countdata <- assay(rse_gene)

replace(countdata,countdata==0,-9)

coldata <- colData(rse_gene)

ddsMat <- DESeqDataSetFromMatrix(countData = countdata, colData = coldata, design = ~ disease.status)

I will be appreciate if any one can help me!!

[cpad0112]

Try this:

library(DESeq2)
load("rse_gene.Rdata")
ddsse=DESeqDataSet(rse_gene,design=~disease.status)

Btw, what is the accession number of the data?

Answer 1

[Kevin Blighe]

Your command using rowSums won't work. What I normally do is something like this:

test <- data.frame(c(1,2,3,4,5), c(1,2,3,4,5), c(1,NA,3,4,5), c(NA,NA,NA,4,5), c(1,2,3,4,5))
colnames(test) <- c("a","b","c","d","e")
test
a b  c  d e
1 1  1 NA 1
2 2 NA NA 2
3 3  3 NA 3
4 4  4  4 4
5 5  5  5 5

test[apply(test, 1, function(x) sum( is.na(x) ))==0,]
a b c d e
4 4 4 4 4
5 5 5 5 5

I originally used this way of filtering for removing transcripts that had zero counts across 5 or more samples, using something like this:

apply(test, 1, function(x) sum(x==0))<5

[e.rempel]

Instead of

test[apply(test, 1, function(x) sum( is.na(x) ))==0,]

you could use the built-in function complete.cases, e.g

test[complete.cases(test), ]

[Kevin Blighe]

Thanks e.rempel - I knew that the function existed but could not remember the name at the time of writing!

Answer 2

[maleknias]

Dear all

I used all your guide but unfortunately all of them have the last same error !!

first:

assay(rse_gene)=assay(rse_gene)[apply(assay(rse_gene), 1, function(x) sum( is.na(x) ))==0,]

dds <- DESeqDataSet(rse_gene, design = ~ disease.status)

second:

assay(rse_gene)=assay(rse_gene)[complete.cases(assay(rse_gene)),]

dds <- DESeqDataSet(rse_gene, design = ~ disease.status)

I will be appreciate if any one can help me!!

[Kevin Blighe]

Try to remove the NA values before you run that function. I'm not sure that the code you have above is going to behave in the way that you expect.

I don't know why you need to use the assay() function before DESeqDataSet()

Can you not just remove NA values from your raw counts matrix before you do anything with DESeq?

[ioannis]

> rse_gene <- data.frame(c(1,2,3,4,5), c(1,2,3,4,5), c(1,NA,3,4,5), c(NA,NA,NA,4,5), c(1,2,3,4,5))
> colnames(rse_gene) <- c("a","b","c","d","e")
> rse_gene[is.na(rse_gene))] <- 0
> rse_gene
  a b c d e
1 1 1 1 0 1
2 2 2 0 0 2
3 3 3 3 0 3
4 4 4 4 4 4
5 5 5 5 5 5
> dds <- DESeqDataSet(rse_gene, design = ~ disease.status)