Question

baseMean differencies according to the model

0

Entering edit mode

jrivera • 0

@jrivera-11162

Last seen 8.4 years ago

Hello!

I would like to understand how does DESeq to compute the baseMean values for each model.

I have tho models:

MODEL A: where design = ~ V1

MODEL B: where design = ~ V2 + V1

When the results matrix appers for each model, there are some bacterias for which baseMean value change between them and the rest arte still the same. Why these changes, and why not in every bacteria?

Tank you!

Deseq2 model • 1.1k views

ADD COMMENT • link 8.4 years ago jrivera • 0

score 0 · Answer 1 · 2016-07-22

0

Entering edit mode

Michael Love 43k

@mikelove

Last seen 22 hours ago

United States

If some very high outlier counts are replaced by DESeq() then the base mean has to be re-estimated, so as to not be influenced by those very high counts. The outlier replacement procedure depends on the experimental design.

ADD COMMENT • link 8.4 years ago Michael Love 43k

score 0 · Answer 2 · 2016-07-22

0

Entering edit mode

jrivera • 0

@jrivera-11162

Last seen 8.4 years ago

Thank you!

Where can I see which individuals have changed their value for certain bacteria?

Following the manual and running until DESeq() function with both designs, counts() of the previous object are the same for both MODEL A and MODEL B.

So, I think that outliers should be replaced in res() function; isn't it?

ADD COMMENT • link 8.4 years ago jrivera • 0

1

Entering edit mode

I'm adding a comment to start a threaded discussion... not that the ADD YOUR ANSWER is for replying to the top post (your original question).

If you take a look at ?DESeq you have information about this, under Details, the paragraph starting with "The argument minReplicatesForReplace is used to decide..."

If you prefer, you can turn off the outlier replacement using minReplicatesForReplace=Inf for DESeq() and turn off filtering with cooksCutoff=FALSE in results() and then examine the genes with high mcols(dds)$maxCooks by eye. These procedures were designed using "standard" RNA-seq datasets and manual inspection may perform better for other types of datasets.

ADD REPLY • link 8.4 years ago Michael Love 43k