HI, this is a bit of odd request.  But is it possible to reassign or combine groups?   

For example say I have  simple group like this: 

group <- c( "a" ,"a", "a" ,"b", "b", "b" )
design <- model.matrix(~0+ factor ( group ) )
fit  <- lmFit( data , design )
contrast.matrix <- makeContrasts(ba = groupb - groupa, levels=design)

However I would like to do this instead. 

group <- factor ( c("a1","a2","a3","b1","b2","b3") ) 
design <- model.matrix(~0+group )
colnames( design ) = group
fit  <- lmFit( data , design )
contrast.matrix <- makeContrasts(ba = (b1+b2+b3) - (a1+a2+a.3), levels=design)

As you can see here every sample is treated like a distinct group but in the contrast I combine them.  I want to do this because under certain genes the groups will be reassign.   However I'm getting an error because each sample is calculated separately.  Is there a way around this?  thanks! 

No, you cannot use limma without replicates. There is no way around that.

You can however use edgeR without replicates -- see the relevant section in the User's Guide.

If you know in advance which genes behave in which ways, then it would be better to analyse the data for each subset of genes separately. That would allow you to use proper groups and use replication.

The other alternative is to simply cluster the genes, for example using a heatmap.