I am looking for co expression pattern of few selected genes in specific cancer vs normal patients samples. It can be either RNA seq or microarray. I found number of databases which claim that they can give co-expression of gene- next bio, geneinvestigator, oncomine and list goes on however most of them either give z score or co-expression p value in one particular data set. Further none of them provide normalized values across different data sets, which may allow me to take that matrix and analyze outside. Even NCBI GEO will allow me to analyze a particular one set at a time. Before I start scanning each data set (which is a lot of work) I was wondering if there is any expression database which can out put normalized expression values across different platforms for normal vs cancer samples.
what you are asking has some limitations -
Having said that, if you say you want to take matrces out, then cross-dataset normalization is not much of a trouble. As you have said already NCBI GEO is one resource, so are these too - EBI ArrayExpress, TCGA. You can get, esp. frm TCGA one combined matrix for all samples, albeit for one cancer type at a time.
For cross-data normalization for removing batch effects, if its microarray then I can suggest SNM. It works well and is from a renowned group.
For RNA-seq I suggest looking into Voom and here How To Transform From Rna-Seq Deseq To Limma Voom() And Makecontrasts