I'm using the Limma package to perform some microarray analysis. In the past I've typically compared 2 groups in contrasts such as "GroupA-GroupB". Occasionally, I've even done comparisons where I've normalized before comparing 2 groups such as "(GroupAExp-GroupAControl)-(GroupBExp-GroupBControl)". In my current analysis I have 4 levels for a particular factor and I'd like to compare 2 of the levels to 2 of the other levels. So, my targets file is as follows.
And basically my contrast is "(A+B)-(C+D)". Which is to say I'd like to find genes that are significantly higher(or lower) in the pooled A and B group compared to the C and D group. Unfortunately, I'm not quite sure I understand the log fold changes that are output. I don't quite understand how Limma is calculating these. I could always just create another column/factor in which I put A & B together as one factor and C & D as another factor but in truth these 4 levels are distinct. So my questions to the group are
1) Is it better to make the contrast as I have or would it be more appropriate to create a new factor and pool A+b into 1 level and C+D into another level.
2) If the contrast I've chosen is correct how does the logFC get calculated, I know that its simply not taking the all the values in A+B and subtracting all the values in C+D as that doesn't yield the same result as what I see in the output.
Any thoughts would be much appreciated.