how merge the gene expression value of two sample

Question

[lkianmehr]

Hello,

I need your help to resolve this question, maybe is easy but I am new about it. I would be appreciated if give me please any idea about that.

actually, I have 8 sample (count file from HTSeq-count of RNA-seq data, with this order: D1_1, D1_2, D3_1, D3_2, R1_1, R1_2, R3_1, R3_2) and make them into 4 group (2 sample in each one). Now I am going to calculate the proportion of gene expression for each group. I have log scaled and normalized them, but I want to merge the ratio of gene expression of each 2 sample together (for ex D1_1 and D1_2) to do the next analysis. could you give me any advice to do that?

my dataset is like this:

                        D1_1                D1_2                D3_1                D3_2    
ENSMUSG00000000001.4    1.39484378430357    2.46452589579488    2.15638312017686    1.39484378430357    
ENSMUSG00000000003.15   0.211843606133756   0.360646924628625   0.307483505135223   0.211843606133756

but I want to have D1 (D1_1 and D1_2) versus D3 (D3_1 and D3_2). I need to merge or don't know should be gotten average or what else?

thanks in advance

[cpad0112]

Good description of the data and problem. Please post some example data and expected output, if possible

[GenoMax]

Is there a reason you are not using an established package like deseq2 to do this analysis?

[lkianmehr]

No, I did differential Gene expression analysis by DESeq2 but now I want to just compare gene type of each group on based of gene expression value, not DGE analysis, you mean I can do this work by DESeq2, if so, how?

[jared.andrews07]

What do you mean by compare gene type? Is there any reason you can't use the DESeq2 logFoldChange values to compare between the groups?

[lkianmehr]

because I want to just calculate the proportion of gene type (Protein coding, miRNA, MiscRNA and etc) on each group, not DGE!

[jared.andrews07]

Still not sure I understand completely. Can you maybe post an example of the output you want/expect?

[lkianmehr]

yes, I expect to have a result like this:

gene    D1  D3  R1  R3  type    
0610005C13Rik   3.26    4.09    3.14    1.03    antisense_RNA   
0610009B22Rik   0.57    0.46    0.91    2.21    protein_coding  
0610009E02Rik   0.77    0.42    0.68    1.31    processed_transcript    
0610009L18Rik   0.58    0.56    4   18.07   bidirectional_promoter_lncRNA

[GenoMax]

So you want to merge (e.g. D1_1 and D1_2 to get just D1, in the example you provide it does not seem to be addition/average) and then annotate each gene row?

[lkianmehr]

No, just to know can I average them to compare them to others as one sample (D1) or not?

[jared.andrews07]

Comparing groups of samples is really best done by something like boxplots, as they give a better indication of the variation within each group. However, with only two samples in each group, boxplots aren't an option, so yes, I suppose taking the average is as good an option as any. I would do it with a grain of salt though, and show the actual sample counts/fold changes in any figures you plan to make.

Answer 1

[h.mon]

Are the D1_1 and D1_2 technical replicates? If so, you can use the function collapseReplicates() from DESeq2, this would sum up the counts. Then you can follow up with your regular analysis.

[jared.andrews07]

Oh yeah, completely forgot about this - it is a good option as well.

[lkianmehr]

yes, they are technical replicates. thanks

[lkianmehr]

Sorry, I need your help to write this command correctly, because I am very new with R and DESeq2, I have written this command and faced with this problem could you please correct command it would be needed?

gr <- factor(c(rep("D1", 2), rep("D3", 2)))
colData <- data.frame(group=gr, type="paired-end")
cds <- DESeqDataSetFromMatrix(cn3, colData, design= ~group)
cds <- DESeq(cds)
cnt <- log2(1+counts(cds, normalized=T))
cntColl <- collapseReplicates(cnt, gr, renameCols = TRUE)

Error in (function (classes, fdef, mtable)  :
  unable to find an inherited method for function ‘assay’ for signature ‘"matrix", "missing"’