Dear Community, 

in an ongoing project i have acquired an updated phenotype information regarding the tumor stage of the cancer samples of one cancer microarray dataset. As you can see from the output below, my samples are paired(each patient has a cancer and its adjucent control sample). For simplicity, in the variable "Tumor_Location" i also included a level="Normal_Tissue" for interpretation purposes:

> head(pData(eset.4))

              Meta_factor  Disease    Location        Study
St_1_WL57A.CEL           0  Normal  left_sided hgu133plus2
St_2_WL57A.CEL           0  Cancer  left_sided hgu133plus2
St_N_EC59A.CEL           0  Normal  left_sided hgu133plus2
St_T_EC59A.CEL           0  Cancer  left_sided hgu133plus2
St_N_EJ31A.CEL           0  Normal right_sided hgu133plus2
St_T_EJ31A.CEL           0  Cancer right_sided hgu133plus2
                Tumor_Stage             
St_1_WL57A.CEL Normal_Tissue         
St_2_WL57A.CEL         T3/T4              
St_N_EC59A.CEL Normal_Tissue       
St_T_EC59A.CEL         T1/T2            
St_N_EJ31A.CEL Normal_Tissue        
St_T_EJ31A.CEL         T3/T4            

Now, i understand that for other types of comparison, like direct comparison of variable Location for cancer samples, i would have to use duplicateCorrelation (and/or other models). As now, my hain interest, is to identify any DE genes between tumors of different stage, for a possible sub-sequent functional enrichment analysis and interpretation, the simplest way to do this is the following ?:

progression <- factor(eset.sel$Tumor_Stage)

> levels(progression)
[1] "Normal_Tissue" "T1/T2"         "T3/T4"      

design <- model.matrix(~0+progression)

fit <- lmFit(eset.4, design)

## and then with makeContrasts obtain the actual comparison of interest:

cm <- makeContrasts(Tumor_stage_comparison= progressionT3/T4-progressionT1/T2, levels=design)..

would be enough? even if i have an imbalance in the group sizes of comparison ? (8 T1/T2 samples & 22 T3/T4 samples)

Thank you in advance !!