I have doubts regarding conducting meta-analysis of differentially expressed genes

Question

[Rohan]

I have generated differential expression gene (DEG) lists separately for multiple OSCC (oral squamous cell carcinoma) datasets, microarray data processed with limma and RNA-Seq data processed with DESeq2. All datasets were obtained from NCBI GEO or ArrayExpress and preprocessed using platform-specific steps. Now, I want to perform a meta-analysis using these DEG lists. I would like to perform separate meta-analysis for the microarray datasets and the RNA seq datasets. What is the best approach to conduct a meta-analysis across these independent DEG results, considering the differences in platforms and that all the individual datasets are from different experiments? What kinds of analysis can be performed?

Answer 1

[LChart]

Since everything is truly independent here (different platforms, different experiments, etc); it seems perfectly acceptable to use Fisher's method for meta-analysis, or if you want to be sign-aware, use a weighted (by sample size) average of inverse-normal Z-scores.