Hello, I'm using DESeq2 to model a multi-factorial experiment with two predictor variables: Genotype (binary) and Passage (integer):
# Make passage variable numeric coldata_exp <- coldata coldata_exp$Passage <- (as.numeric(coldata_exp$Passage)) dds_exp <- DESeqDataSetFromMatrix(countData = cts, colData = coldata_exp, design= ~Genotype*Passage) keep <- rowSums(counts(dds_exp)) >= 10 dds_exp <- dds_exp[keep,] # Make the model, perform likelihood ratio test dds_exp <- DESeq(dds_exp, test="LRT",reduced = ~ Genotype + Passage)
We would like to derive some measure of 'effect size' of the interaction which is that is comparable between genes; or, how much does v1 affect gene expression in a v2-dependent manner? We have arrived at the standardized Beta coefficient (Wikipedia) as an appropriate measure of effect size.
My few questions are: Is it appropriate to compare standardized Beta values across genes? Does the DESeq2 coef() function output raw regression coefficients, or are they standardized Betas?
Thank you! Owen