Hello everyone,

I am working with three microarray datasets from different platforms from GEO to investigate gene expression patterns in specific cell types. Within each dataset, I have carefully selected a subset of samples that meet my specific criteria (e.g., cell type, treatment conditions).

My main question is: what is the best approach to normalize these selected samples across the different platforms? My goal is to minimize technical variation while preserving the biological differences I am interested in.

Furthermore, I am wondering whether it is better to run PCA on each dataset separately or combine the selected samples into a unified dataset and then run PCA.

Adding to my questions, I am also puzzled by the fact that these datasets use different annotation platforms. This leaves me unsure about the best way to bring these datasets together for differential gene expression (DEG) analysis.

Should I try to re-annotate everything to a common platform or are there other strategies that might be more effective?

Any advice or insights based on your experience would be much appreciated!