statistical test on count data for differential expression

Question

[tonja.r]

I have two conditions: treated and untreated. Each condition hast two biological replicates. I also have 7 different features represented as count data. I need to find differentially expressed genes.

If I simplify the model till only one replicate per condition, I could do exact Fisher Test or clustering.

Are there some more methods to test for differential expression that I am forgetting? And how could I integrate the second replicate?

CONDITION   GENE    FEATURE1    FEATURE2    FEATURE3    FEATURE...
Treated 1   gene A  3432        123             45454   
Treated 2   gene A  4534        234             46567   
Untreated 1 gene A  124         134             2345    
Untreated 2 gene A  1245        34              2345

[cpad0112]

DESeq or DESeq2? DESeq2 also accepts counts matrix.

[michael.ante]

Hi Tonja,

What are these features?

Cheers,
Michael

[tonja.r]

That are the number of reads in specified regions of the genome coming from ChIP-seq from different histones.

Answer 1

[Adrian Pelin]

I think DESeq has all you need https://bioconductor.org/packages/release/bioc/html/DESeq.html

It even accepts a matrix of counts as input.

[tonja.r]

As far as I understood it takes only one feature. Columns are the samples with the genes as rows. With the example above it would be something like this:

        Treated 1  Treated 2  Untreated 1  Untreated 2
gene A  3432       4534       124          1245

So, I guess when you suggested to use DESeq, you meant to transfer my 8 features into samples. So, that I will have 4*8 samples in fact.But I do not think it would be correct to do that.

        Treated 1/Feature 1  Treated 2/Feature 1  Untreated 1/Feature 1  Untreated 2/Feature 1  Treated 1 /Feature 2  Treated 2 /Feature 2  Untreated 1 /Feature 2
gene A  3432                 4534                 124                    1245                   123                   234                   134