Is Batch correction required for the same micro array platform?

Question

[au.rinki.bio]

Hii I am handling microarray data and I have two datasets with the same platform (GPL750) one of which contains tumor samples and another one has normal samples. I want to merge these two datasets for differential expression analysis. Is it required batch correction?

[Buffo]

Before to think about batch correction I suggest considering this: Do you have replicates? How dispersed they are? what are the causes of the batch effect? Do I have the batch effect?. If you don't have replicates, Is it possible to have the batch effect without replicates?

Answer 1

[Benn]

If the data is from two different experiments, you should correct for batch effect indeed. It sounds however, that you have an impossible design: tumors in one batch and normal in the other batch. It is not possible to correct for batch in this case, since there is no overlap between the batches. You need tumors and/or normals in both batches.

[au.rinki.bio]

sorry for the very basic question can I consider normal and case sample from different dataset? is it valid for differential expression analysis?

[Benn]

No that would not be wise and sound, since you can't distinguish between real effect (between normal and tumor) and between batch effect.

[au.rinki.bio]

In this case how to remove batch effect?

[Benn]

Sorry, not possible.