normalized input data for GAGE (R) - log-transformed or not?

Question

[AdrijaK]

Does GAGE package in R require log-transformed normalized expression values or "just" normalized expression values as an input?

Answer 1

[informatics bot]

What are you using GAGE for (RNA-seq or microarray)?

For RNA-seq you can use log2 normalized data:

cnts=raw_gene_count_data
sel.rn=rowSums(cnts) != 0
cnts=cnts[sel.rn,]
libsizes=colSums(cnts)
size.factor=libsizes/exp(mean(log(libsizes)))
cnts.norm=t(t(cnts)/size.factor)
cnts.norm=log2(cnts.norm+8)

For microarray data, they use RMA/FARMS normalization in the vignette.

[AdrijaK]

Thank you. Currently my input is RMA normalised microarray data. So I guess I can analyse this directly.

[Rob]

Why does the last line add 8 to cnts.norm? I understand something must be added to prevent -infs in the log2 transformed dataset, but why specifically 8?

Answer 2

[bigmawen]

You may use RMA, FARMS or other normalization methods. Array normalization usually does log(2) transformation too like in RMA and FARMS. If not, it is always advisable to do log2 or log transformation on array or RNA-Seq data for differential expression or pathway analysis.