Hi,
I am working with 8 different microarray datasets comprising both Affymetrix and Illumina arrays to perform Meta-Analysis or Cross-Platform Analysis and identify biomarker signatures between disease vs control condition.
I downloaded each dataset (*.soft file from NCBI GEO dataset) individually and then sequentially processed by applying "Quantile Normalization" followed by Probe to Gene Mapping. The final dataframe/matrix will contain "Gene Symbols" as rows and "Samples" as columns for each dataset. Hence, all the 8 individual datasets were processed similarly irrespective of the array platform.
I have a question, since, the *.soft/series matrix file from NCBI GEO contains already normalized/log2 normalized data (all 8 datasets of interest were normalized/log2 normalized). Please let me know if it makes sense to re-normalize the data again using "Quantile Normalization" to accomplish meta-analysis or cross-platform analysis?
Thank you,
Toufiq
Hello, I'm not sure if you want to perform normalization on top of the already normalized data. Generally, this practice is not followed. Consider that these data might have been normalized with different methods from study to study. Although I haven't performed metanalysis in the past, might be worth checking the approach we took in our article: https://pubmed.ncbi.nlm.nih.gov/29186820/
In summary, we just reanalysed each dataset and integrated the functional enrichment analysis results.