Dear Phụng Dương

1. If these datasets were generated by different platforms, you can try using this guide "Cross-Platform Meta Analysis" = https://bioconductor.org/packages/devel/bioc/vignettes/crossmeta/inst/doc/crossmeta-vignette.html

2. If these datasets were generated by an equal platform, you can split in control group and disease group and do a batch correction of microarray dataset

https://www.biostars.org/p/270756/

https://www.biostars.org/p/266507/

https://www.biostars.org/p/121119/

https://www.biostars.org/p/101283/

Article Removing Batch Effects in Analysis of Expression Microarray ...

Article Batch correction of microarray data substantially improves t...

3. You can analyze each dataset individually and overlap the DEGs. For a DEG be considered "differentially expressed" this DEG must be present in at least 4 of the 9 datasets and that follow the same direction of expression (Up or down)

Article Meta-analysis of host response networks identifies a common ...

Thank you for your suggestion. It is valuable and useful materials to carry out meta-analysis. But I wonder how should I calculate fold change for each genes after I have finished meta-analysis? Now I have about more than 1000 DEG. It is easy to calculate foldchange for 1 dataset, but what about the overall fold change for the gene over 9 datasets? I think the sample size of each dataset is not similar together so it is not possible to calculate the mean of foldchange among dataset. Could you please give me some suggestion for this?

I would rather do a p-value based meta-analysis weighted by sample size. expecially when the platforms and the models are different, it gives less biases.