# Experimental Design for Differential expreression analysis

I have a Normal esophageal Fibroblasts (NOFs) cultured in DMEM media; The same NOF also have been cultured with a tumor sample from a patient named 005 on DMEM media; I have also Cancer Associated Fibroblasts (CAFs) cultured on DMEM, on DMEM + tumor from a patient named as 005 and DMEM + tumor from a patient named 036 something like this picture;

I want to know if NOF and CAF are different in terms of gene expression but I am not sure about design for DESeq2 or Edger and ANOVA.

By @Devon's help I have done so

> mycols
cell_type co_cultured
G2       NOF          NO
G3       NOF         YES
G4       CAF          NO
G5       CAF         YES
G6       CAF         YES

dds=DESeqDataSetFromMatrix(countData = NOFCAF,colData = mycols, design        =~ cell_type*co_cultured)


Now, I am wondering what would be a suitable contrast for results function in DESeq2

> resultsNames(dds)
[1] "Intercept"                   "cell_type_NOF_vs_CAF"        "co_cultured_YES_vs_NO"
[4] "cell_typeNOF.co_culturedYES"

> res=results(dds)
> res
log2 fold change (MLE): cell typeNOF.co culturedYES
Wald test p-value: cell typeNOF.co culturedYES
DataFrame with 2545 rows and 6 columns


Thanks a lot for correcting me

• Do you have replicates of any of these? – Devon Ryan Jan 7 '19 at 4:53
• Following Devon comment, you don't need biological replicates, technical replicates are worthy. It would help to know how many samples do you have available of each type. In addition, it is important to consider potential batch effects (where all samples extracted and sequenced at once?) and con-founders (are some sample from older patients? Is sex equally distributed) – llrs Jan 7 '19 at 8:18
• No, you really need biological replicates. The conclusions you can draw from technical replicates only is limited. – swbarnes2 Jan 7 '19 at 17:11
• You can figure out that error from DESeqDataSetFromMatrix yourself, you shouldn't need any help (hint: design is stated incorrectly). – Devon Ryan Jan 8 '19 at 11:22
• You don't require any further help on this, you need to fly free a bit more. – Devon Ryan Jan 8 '19 at 12:19

You can use a design of ~cellType * coculture where cellType has levels NOF and CAF and coculture has levels yes and no (ideally you'd block by patient, but that's not an option for you).