How to get fitted per sample means and dispersion from DESeq2?

Question

[kjkjindal]

Hi,

DESeq uses the following model for differential expression analysis:

K_ij ~ NB(mu_ij, alpha_i)

mu_ij = s_j q_ij

log2(q_ij) = x_j. beta_i

Is there a way to access the specific 'mu_ij' and 'alpha_i' values?

I have been looking around and found assays(dds)[['mu']] giving something like mu_ij but am not sure if those are indeed the values that I am looking for.

Thanks!

Answer 1

[Kevin Blighe]

Take a look at the Access to all calculated values of the tutorial Analyzing RNA-seq data with DESeq2.

Dispersion is calculated on a gene-wise basis and the final modelled dispersion can be obtained with mcols(dds)$dispersion or dispersions(dds). The estimated gene dispersions that go into the Bayesian model as priors can be obtained with mcols(dds)$dispGeneEst

For the mean, it states:

[kjkjindal]

Hi, that worked great. Thanks a lot Kevin!

[Kevin Blighe]

You're welcome - no problem.