Hi

DESeq is a great tool. I use it a lot. I am trying to figure out how to use it more skillfully. I wonder if I would get better results by changing my design? If I use the linear model function in [R] I get the coefficient of determination or R². Is there a way to get that value for each gene?

Maybe there is a better way to evaluate the fit?

Kind regards

Andy

p.s. there are lots of ways to model biological systems. Can anyone point me to a good reference?

You can compare the estimated values ("mu", see "Access to all calculated values" in the vignette) to the observed counts (counts(dds) without scaling). Note that by adding more covariates, the estimated values will become closer to observed counts (hence the LRT for assessing if the improvement is more than what would be expected if the true relationship was null).