I have a microarray data with many duplicated probe names (some are duplicated 2 times, some 10 times, some 35 times). The package limma offers a function avereps() to average expression across the same probe names, e.g. y.ave <- avereps(y, ID=y$genes$ProbeName). However, I do not want to run lmFit on averaged expression levels.

A piece of my data:

             ProbeNames      Slide1.1_1 Slide1.1_2 Slide1.1_3 Slide1.1_4
1           A_23_P122775   3.380455   3.428704   3.665035   3.594631
2           A_23_P122775   3.439826   3.428704   3.351302   3.291165
3           A_23_P122775   2.991987   3.336501   3.531051   3.594631
4           A_23_P122775   3.688537   2.869843   3.531051   3.720810
5           A_23_P122775   3.156077   3.467261   3.531051   3.594631
6           A_23_P122775   3.926539   3.036624   3.665035   3.431492

Now, my model looks like fit <- lmFit(~ slide + time). There is no interaction between slide and time. However, with that model I obtain estimates and p-values for each row (for each duplicate). Is there any option to run lmFit incorporating an information about duplicated probe names? Thanks in advance for help.