DEXSeq with experimental variables

[adumitri]

Hi,

Can anyone let me know if the formulae displayed below are accurate when trying to account for two different experimental variables (batch and age) with DEXSeq? Also, what does the "I(exon == exonID)" in the regression model do? Thank you!

formuladispersion <- count ~ sample + (condition + batch + age) * exon

--> formuladispersion used with the estimateDispersions() function

formula0 <- count ~ sample + batch * exon + age * exon + condition
formula1 <- count ~ sample + batch * exon + age * exon + condition * I (exon == exonID)

--> formula0 and formula1 used with the testForDEU() function

[areyes]

Hi @adimitri,

If I understood, you want to see differences between conditions using as blocking factors the batch and age of your samples right? If that is the case, your dispersion formula looks good, and your formulas for testing should be:

formula0 <- count ~ sample + (batch + age) * exon + condition
formula1 <- count ~ sample + (batch + age) * exon + condition * I(exon == exonID)

the I(exon == exonID) is a way to specify the contrast with a specific exon, if you have a look inside the testGeneForDEU function, you will see that exonID is a variable that iterates through each exon id from each gene.

Alejandro

[adumitri]

Hi Alejandro,

Thank you for your answer - this is exactly what I was looking for!

Alexandra