Linear Model Design Formula for "Complex" Experimental design
1
0
Entering edit mode
5.4 years ago

Hello,

I have a question about creating the "design formula" for my linear model within edgeR.

First I will explain the experimental design:

7 subjects donated tissue samples.

For each donor, their tissue was separated in 4 groups (n = 28):

  • 1st group was stimulated with an enzyme for 5 days.
  • 2nd group was stimulated with the same enzyme for 10 days
  • 3rd group  was a mock-treatment control group for 5 days
  • 4th group was a mock-treatment control group for 10 days

The top of the design matrix looks like this:

sample treatment time
sample1   enzyme   5.day
sample1   mock       5.day
sample1   enzyme   10.day
sample1   mock       10.day
sample2   enzyme   5.day
sample2   mock       5.day

 

My design formula looks like this:

model.matrix(~sample+time+treatment, design)

But I'm not sure it is correct...

Please help,

Lando

EDIT:

The main question we're asking is:

1.) Which genes are differentially expressed between treatment (enzyme stimulation) and control?

Sub-question:

2.) does 10 day stimulation significantly differ from 5 day stimulation, when compared to control?

 

p.s. I will actually upvote your answer if you help me solve this :D

limma edgeR RNA-Seq linear model R • 2.2k views
ADD COMMENT
0
Entering edit mode

What question do you want to answer? That's what ends up determining the design (though what you're using is likely more or less what you'll end up wanting).

ADD REPLY
0
Entering edit mode

The main question we're asking is:

  1. Which genes are differentially expressed between treatment (enzyme stimulation) and control?

    Sub-question:

    • does 10 day stimulation significantly differ from 5 day stimulation, when compared to control?
ADD REPLY
0
Entering edit mode

I think you can drop the "sample" variable in the model, unless you are interested in testing whether there are differences between the samples themselves.

ADD REPLY
1
Entering edit mode

These are human samples, so I wouldn't recommend that.

ADD REPLY
4
Entering edit mode
5.4 years ago

Given your comment, you want

model.matrix(~sample+time*treatment, design)

The coefficients of interest are treatment and time:treatment (the interaction). This will still control for a time and sample effect.

ADD COMMENT
0
Entering edit mode

Awesome, I'm looking at the model matrix and it looks right.

I really appreciate it!

:D

ADD REPLY

Login before adding your answer.

Traffic: 1662 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6