how to normalize RNA-seq with microarray data?

[zhaopeihua]

hi,

I have some gene expression data measured by RNA-seq or microarray. but RNA-seq data gene expression level generally higher than microarray.How should I normalize the data to find the different expression gene?

Thanks in advance
Any help would be appreciated.

[dpryan]

Most of the methods in use aren't meant for this in any way. However, you might be able to use WGCNA, which can work on microarrays or RNAseq, just include the technology type as a factor. Similarly, perhaps limma will work OK (again, include the technology type as a batch effect).

This assumes, of course, that your groups are not separated according the the techonology (i,e, group A has microarray data and group B RNAseq any you want to find DE genes between the two). That's unlikely to work well regardless of the method.

[DonDolowy]

If I want to compare my RNA-seq with publicly available microarray datasets, I tend to work with fold expression instead of actual expression values. That should make the direct comparison legit.

[dpryan]

That won't make the comparisons legitimate if you actually plan on looking for DE genes between the technologies. The signals on a microarray are independent, whereas read counts are competitive, so they have different distributions. You can try to normalize that away but I'd be hesitant to believe the results without having first done a validation experiment with known results (e.g. running the same samples on both platforms and ensuring that the resulting profiles are the close enough and there are effectively no DE genes in a comparison).

[zhaopeihua]

If I want to compare my RNA-seq with publicly available microarray datasets, I tend to work with fold expression instead of actual expression values. That should make the direct comparison legit.

What is the meaning of fold expression?