Question

Deseq2: 3 factors design

0

Entering edit mode

marburg2107 • 0

@marburg2107-14646

Last seen 6.4 years ago

Many people were confused to handle three-factors with DESeq2 though many posts on this website.

Three factors are temperature(A, B); water content (C, D); co2 concentration (E, F).

I want to know the following effects: (1)temperature, (2)water, (3)co2, (4)temperatureco2, (5)waterco2, (6)temperaturewaterco2, (7)temperature*water.

DESeqDataSetFromMatrix(countData = countData,colData = colData,design = ~co2+warm+drought+co2:warm+co2:drought+warm:drought+co2:warm:drought) --------------Error in checkFullRank(modelMatrix)

My R version is 3.4.2 (2017-09-28), Deseq2 version is 1.18.1

deseq2 multiple factor design • 1.5k views

ADD COMMENT • link updated 6.4 years ago by Michael Love 41k • written 6.4 years ago by marburg2107 • 0

score 0 · Answer 1 · 2017-12-17

0

Entering edit mode

Michael Love 41k

@mikelove

Last seen 1 minute ago

United States

I notice you posted the same question simultaneously to Biostars, this is discouraged as it duplicates the effort or volunteers.

There are many valid formula for when you have three factors with potential interactions. There is not a specific software question or reproducible example here. I recommend that you meet with and collaborate with a statistician who can help you setup an appropriate design and set of statistical tests. You can use any linear model formula with DESeq2.

ADD COMMENT • link 6.4 years ago Michael Love 41k

0

Entering edit mode

Hi, Michael. Thanks for your time!

However, it's really difficult to find a statistician around us.

Total 72 samples, 6 treatment,12 replicates.

6 treatments were: co2-400ppm, co2-400ppm+drought, co2-400ppm+drought+warm, co2-600ppm, co2-600ppm+drought,co2-600ppm+drought+warm

Maybe you can tell me more about the statistical tests.

ADD REPLY • link 6.4 years ago marburg2107 • 0

0

Entering edit mode

I’m unfortunately too busy right now to give personal feedback on this analysis. For a design where second order interactions are possible, figuring out what kind of modeling choices to make, which interactions to include, how to construct and interpret the results is not trivial, and the analyst should have a background in linear modeling in R.

ADD REPLY • link 6.4 years ago Michael Love 41k