Combining in vivo and in vitro studies are not generally recommended, but it depends on you goal and problem. About MAMA package please clarify your question. I may be able to help. I recommend MetaOmics. Have a look, too. 

http://www.pitt.edu/~tsengweb/MetaOmicsHome.htm

 Many thanks dear Dr. Jamalkandi for your reply. Although I dont have any experience with meta-analysis of high throughput data like microarray and RNA-seq, I have good knowledge of analysis of these type of data with R software. That's why I was thinking of MAMA package in R. I will check MetaOmics. Thanks for your recommendation.

Best,

MetaOmics is also R-based package. Generally there are two kind of microarray meta-analytics: 1- probe level, 2- differential gene level. MAMA and metaomics are for the 2nd case. If you want to have a probe level meta-analysis your choice will be Virtualarray. 

"Key issues in conducting a meta-analysis of gene expression microarray datasets" is a very useful paper for met-analysis of microarray datasets. 

Thank you. Your reply was so informative for me. I have 12 data sets with different platforms, affymetrix, agilent, illumina for RNA-seq. As you see, some of them are probe-level and others are gene-level. In this situation,  I will have problem?

I think it is better to find your DEGs for each study. The combine the results based on their effect sizes, p-values, votes and etc. There are a list and related references in MAMA tutorial. This was for microarray datasets. Also you can use VirtualArray (R package) for probe level meta-analysis of different platforms and technologies. But you will need a bit more and complex statistical considerations. 

Use RNA-seq data for cross validation. 

Bests

 Then I should check MAMA,VirtualArray and MetaOmics. By limma in R and CLC software, I have found DEGs for all platforms. In MAMA, there are lots of methods for meta-analysis, combining p-values, effect sizes and... Can I do a meta-analysis on P-values to find significant genes and after that doing a meta-analysis on effect sizes of those significant genes to find up and down regulated genes? 

Limma gives you DEGs. Each deg has a p value and a fold change. If you want to use p-value combination in MAMA, you should use limma output. I mean limma output will be MAMA input data. Therefore, you can select one of the meta-analysis strategies in MAMA. If you are not sure about the strategy you need, I recommend the following paper before performing your met-analysis:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3351145/

Thank you Dr. Azimzadeh, sorry to bother you with my naive questions. I have list of DEGs with p-value, fold change and another statistical parameters. Since there are different platforms, I have different number of genes in data sets. I think I should select a good subset of these genes (or probes)!!!. It is better to study these recommended papers, after that I will return you. 

Thank you for your time and consideration.

You are welcome. You need to select shared genes among different platforms. Unfortunately, you should discard unique genes within each platform. Also probe names are different, each platform has several probe names which fit to one gene or vice versa. These are difficulties of cross platform meta-analysis. 

There were a lot of methods for meta analysis in micro array data sets. It is possible except R software, you can use each method by other software. but the performance of meta-analysis methods should assessed via ROC curves and prediction accuracy.