How To Use Present/Absent Calls Based Filtering Of Probes?

Question

[Diwan]

Hi,

I am analyzing the affy microarray data with >10 conditions in triplicates. I would like to know, what is the cut-off or the best way to use MAS5 present/absent calls.

Can I just remove probes that are absent in all conditions?

In R, I do:

calls <- mas5calls(eset1);  
calls <- exprs(calls);
absent <- rowSums(calls == 'A');

Or is there any other more commonly used cut-off for the present/absent calls?

Thanks in advance.

Diwan

Answer 1

[seidel]

If you want to use MAS5, removing probes marked absent is common. And a straightforward approach is to include only probes that are present under all conditions. However, you should consider the case in which some genes that may be down-regulated under one of your conditions may go from being detectable (Present) to undetectable (Absent). If you do your analysis and only exclude genes that are marked Absent under all conditions, you should be able to par down your data set, and still retain genes at the interface.

Another option is to simply use an RMA based approach, not worry about the Presence/Absence conundrum, and let the variance tell you who to pay attention to. The data can look different between RMA and MAS5 treatments, and both have merits that have been discussed in the literature. The overall biological message shouldn't depend on one or the other, but both can be informative.