DESEeq differential analysis after PEER normalization

Question

[tonja.r]

I want to use PEER normalization to run DESeq on my data. RUVseq outputs factors of unwanted variance and they can be introduced into the design matrix. However, PEER, as far as I understood, outputs the normalized data already:

Yc = PEER_getResiduals(model)

Is there a possibility to get those factors to insert them into design matrix for analysis with DESeq?

Answer 1

[informatics bot]

Before you add the factors to the design matrix, you should first retrieve the weighted values for each gene within each factor using:

weights = PEER_getW(model)

You NEED TO KNOW what you are biologically factoring into your linear model before you blindly run DESeq on PEER factors. To ascertain the biological significance of each factor, use DAVID on the top weighted genes. If the weights have a normal distribution you can take the highest and lowest 2.5%, if the distribution is right skew take the top 5%, and run them through DAVID. Once you verify the biological basis for your peer factors, then you can get the factor values (for the DESeq design matrix) using:

factors = PEER_getX(model)