Hello,

I am trying to analyze GSE7864 and would like to know how miR34a, miR34b, and miR34c influence the gene expression, i.e., what is the Differentially  expressed genes (DGE) caused by miR34a, miR34b, and miR34c, respectively?

The following is my code, but I am not sure how to construct design matrix according the  tTarget  information (i.e., targets frame according to Limma tutorial). I am trying to select a subset according to different Cy3 and the subsetted targets frame called sTarget, I know sTarget belongs to two-color with common reference designs (p37 in Limma tutorial), but using sTargets only can not build linear model in Limma since no enough replicates for each treatment. In this case, how can I get the DGE permuted by miR34a, miR34b, and miR34c, respectively? Or is there anther way to obtain the DGE by using all arrays instead of just 3 like in sTargets? If so, how to contrast the design matrix and contrast matrix? I can not find the similar examples in Limma tutorial.

If 2-fold change is used the measure the extent of DGE for GSM190752 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM190752), then FC = 10^VAUE (since VALUE is LOG10 RATIO)? and the genes with abs(FC) > 2 are DGE permuted by miR34a? 

Any help is appreciated!

Kevin

# https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE7864

eset <- getGEO(filename = "GSE7864_series_matrix.txt.gz")

tCy3 = rep(c("A549H1", "HCT116Dicer", "TOV21GH1", "DLDDicer", "HeLa", "A549p53", "TOV21Gp53"), each = 4)

tCy5 = rep(c("Luc", "miR34a", "miR34b", "miR34c"), times = 7)

pd <- pData(eset)

tTarget <- data.frame(gsm = rownames(pd), Cy3 = tCy3, Cy5 = tCy5)

###########################################################

> tTarget
         gsm         Cy3    Cy5
1  GSM190751      A549H1    Luc
2  GSM190752      A549H1 miR34a
3  GSM190753      A549H1 miR34b
4  GSM190754      A549H1 miR34c
5  GSM190755 HCT116Dicer    Luc
6  GSM190756 HCT116Dicer miR34a
7  GSM190757 HCT116Dicer miR34b
8  GSM190758 HCT116Dicer miR34c
9  GSM190759    TOV21GH1    Luc
10 GSM190760    TOV21GH1 miR34a
11 GSM190761    TOV21GH1 miR34b
12 GSM190762    TOV21GH1 miR34c
13 GSM190763    DLDDicer    Luc
14 GSM190764    DLDDicer miR34a
15 GSM190765    DLDDicer miR34b
16 GSM190766    DLDDicer miR34c
17 GSM190767        HeLa    Luc
18 GSM190768        HeLa miR34a
19 GSM190769        HeLa miR34b
20 GSM190770        HeLa miR34c
21 GSM190771     A549p53    Luc
22 GSM190772     A549p53 miR34a
23 GSM190773     A549p53 miR34b
24 GSM190774     A549p53 miR34c
25 GSM190775   TOV21Gp53    Luc
26 GSM190776   TOV21Gp53 miR34a
27 GSM190777   TOV21Gp53 miR34b
28 GSM190778   TOV21Gp53 miR34c

###########################################

# selected Cy3 and Cy5

sCy3 = c("A549H1")
sCy5 = c("miR34a", "miR34b", "miR34c")

isSelected <- (tTarget$Cy3 %in% sCy3) & (tTarget$Cy5 %in% sCy5)

sTarget <- tTarget[isSelected, ]

###########################################

> sTarget
        gsm    Cy3    Cy5
2 GSM190752 A549H1 miR34a
3 GSM190753 A549H1 miR34b
4 GSM190754 A549H1 miR34c

You can't analyse this data using limma because there is no replication.

If you want to find probes with a 2-fold change in any particular sample, sample j say, simply select probes with abs(exprs(eset)[,j]) > log10(2).