I want to do some meta analysis using GEO datasets, it contain different PlatForms microarray data and RNASeq data.(different PlatForms, contain different probes,) I have two pre-processing procedure, as following, procedure1: 1. probes to symbol in each GSEs, 2. limma::voom(RNASeq_data), change it became microarray distribution.(the RNASeq data is counts) 3. intersect each GSEs' and RNASeq symbols.(remain about 8000+gene symbols) 4. with the same symbol numbers and same distribution. I perform the GSVA score.
procedure2: about the ssGSEA method , use RNASeq(TPM), no need to change the RNAseq data's distribution, but only intersect each GSEs' and RNASeq symbols, and perform the ssGSEA score.
my questions are :
are those two procedures right?
and is intersect each GSEs' and RNASeq symbols necessary？ because ssGSEA have its normalization(NES),
and I think GSVA is not suitable for this condition, because of the beach effect between different platform, GSVA maybe can change the genes' rank of the metadata.
