Hi everyone,

I have an expressionSet eset containing 12 samples & 21000 genes of cardiac tissue from Mouse. The samples' information is as follows:

                               chip sampleid   label strain treat       batch
   4769_48479_WTGFP_MoGene2.0st.CEL    48479   WTGFP     WT   GFP           1
  4769_48480_WTCRE1_MoGene2.0st.CEL    48480   WTCRE     WT   CRE           1
  4769_48481_WTCRE2_MoGene2.0st.CEL    48481   WTCRE     WT   CRE           1
  4769_48483_WTCRE4_MoGene2.0st.CEL    48483   WTCRE     WT   CRE           1
4769_48484_P1P4GFP1_MoGene2.0st.CEL    48484 P1P4GFP    DKO   GFP           1
4769_48485_P1P4GFP2_MoGene2.0st.CEL    48485 P1P4GFP    DKO   GFP           1
4769_48486_P1P4GFP3_MoGene2.0st.CEL    48486 P1P4GFP    DKO   GFP           1
4769_48487_P1P4CRE1_MoGene2.0st.CEL    48487 P1P4CRE    DKO   CRE           1
4769_48488_P1P4CRE2_MoGene2.0st.CEL    48488 P1P4CRE    DKO   CRE           1
4769_48490_P1P4CRE4_MoGene2.0st.CEL    48490 P1P4CRE    DKO   CRE           1
   4922_52242_WTGFP_MoGene2.0st.CEL    52242   WTGFP     WT   GFP           2
   4922_52243_WTGFP_MoGene2.0st.CEL    52243   WTGFP     WT   GFP           2

As you can see there are two batches, the first 10 samples belong to batch 1 and last two samples belong to batch 2. I tried to batch correct the samples using:

pheno <- pData(eset) #get the phenotypic data from eset (expressionSet)

pheno$batch <- c(rep(1,10),rep(2,2)) #define batches

expr.test <- exprs(eset) #extract expression data into a double matrix

mod = model.matrix(~1, data=pheno) #Model matrix for other covariates besides batch

combat.expr<-ComBat(expr.test,pheno$batch,mod)

I did my PCA on this but doesn't look right. I think it is because in the samples with label WTGFP, one comes from batch 1 and other two come from batch 2.

How can I incorporate this information in my model.matrix function so that it corrects for different batches but does not differentiate between similarly labeled groups?