Question

RNA-Seq and edgeR: remarkable differences between cpms and fitted values

0

Entering edit mode

David R ▴ 90

@david-rengel-6321

Last seen 10 months ago

European Union

Hi, I am treating NovaSeq-produced RNA-Seq data with edgeR. In the analysis I look at what I call “strain” effect. There are DEGs for this factor that I find are biologically hard to explain. In the attached file, there are two examples. It is easy to see why they are declared as differentially expressed since fitted values vary among strains (right hand plots in the figure). What is hard for me to follow is how those fitted values might be obtained when the TMM-normalized cpms do not reveal that behavior (left-hand plots). In the latter case, cpms tend to 0 except for one repeat on the WT strain. I thank you in advance for any help/comment on the matter. Best, David

fitted vs cpm expression

RNA-Seq edger fitted values differential expression normalization • 1.6k views

ADD COMMENT • link updated 5.4 years ago by Gordon Smyth 51k • written 5.4 years ago by David R ▴ 90

0

Entering edit mode

From the FAQ

How do I put images into my posts?

You can also put images and plots into your posts. All you need to do is create (or login to) a free account with http://imgur.com and upload any images or plots that you want to share. Then you can enter the imgur generated 'Direct Link' url into the 'URL' location that is available when you click on the 'image' button in the text editor. The plot will be embedded right into your post. To "attach" other files to your posts, upload them to Dropbox or another similar site and include the URL in your post. Make sure the URL is publicly accessible.

ADD REPLY • link 5.4 years ago James W. MacDonald 67k

0

Entering edit mode

Hi James, Thanks, but is not it what I have done? I mean, there is a link to the figure. I've put a new imgur link just in case the previous one was not public. Hope it works. Best, David

ADD REPLY • link 5.4 years ago David R ▴ 90

0

Entering edit mode

Edit I can't get the 'direct link' from your post, so it doesn't work for me.

ADD REPLY • link 5.4 years ago James W. MacDonald 67k

0

Entering edit mode

I can get a direct link to one of my own imgur posts, and it looks like this

enter image description here

ADD REPLY • link 5.4 years ago James W. MacDonald 67k

0

Entering edit mode

I updated this post -- was the first time I used imgur, actually, so wasn't sure what URL hacking was necessary to get the image to show up.

You don't copy & paste the URL of the image page from imgur (which is https://imgur.com/a/oxlZBoo), but rather you have to right click on the image itself and get that URL (which is https://i.imgur.com/daGn4Q0.png).

I guess having to do this is obvious in retrospect, but still a bit of a PITA for most people the first time or so.

From the text in the FAQ, it's not clear where one gets the "Direct Link" that's written there ... I'm guessing that this is something that imgur shows when the image is first uploaded/created? Because I can't find it on the imgur webpage after the fact ...

... aaaaaaannnnnnyyyyyyyway ... there we have it.

ADD REPLY • link 5.4 years ago Steve Lianoglou ★ 13k

0

Entering edit mode

Yes, but if it's your imgur post, and you click on it, you are presented with a dialog box that looks like this:

enter image description here

And if you look to the right, there's a Direct Link that you can copy to post. Hence the part in the FAQ about a 'Direct Link'.

ADD REPLY • link 5.4 years ago James W. MacDonald 67k

0

Entering edit mode

recursion

ADD REPLY • link 5.4 years ago Steve Lianoglou ★ 13k

0

Entering edit mode

perfect, thanks for the tips! :-)

ADD REPLY • link 5.4 years ago Steve Lianoglou ★ 13k

1

Entering edit mode

Keymaker, my friend

enter image description here

ADD REPLY • link 5.4 years ago James W. MacDonald 67k

0

Entering edit mode

Can you add more details (code) to your question, please?

Can you share a bit of your $samples data.frame and show us the code you used to generate the design and the fit?
Where did you get these fitted values from? Is it off of some $fitted.values somewhere?

ADD REPLY • link 5.4 years ago Steve Lianoglou ★ 13k

score 0 · Answer 1 · 2019-06-13

0

Entering edit mode

Gordon Smyth 51k

@gordon-smyth

Last seen 5 hours ago

WEHI, Melbourne, Australia

It doesn't make any sense to plot fitted values because they are proportional to the library sizes, which in turn vary independently of the groups or treatment conditions of interest. The reason why we recommend that you use CPMs or logCPMs for plotting purposes is that the library sizes have been divided out.