Search
Question: DESeq2 for multiple comparisons over time
0
gravatar for ucheuna
3 months ago by
ucheuna0
ucheuna0 wrote:

I wish to know genes that are significantly expressed between knockout and wild type mice as they age from 3 months, 6 months, 12months and 24months. I wonder the best way to analyze such data. I combined Age and condition (genotype) and performed DESeq2 as follows: 

> coldata$Age_condition <- factor(paste0(coldata$Age, "-", coldata$condition))

> dds <-DESeqDataSetFromMatrix(countData = countdata, colData = coldata, design = ~ Age_condition)

> dds

> dds <- DESeq(dds)

I am stuck at this point and having issues extracting differential expression results to see genes that significantly differ between 3-6, 6-12, 12-24, 3-24, 3-12, 6-24 months, upon knockdown of the gene. What might be the best way to go about this? Thank you

 

 

 

ADD COMMENTlink modified 3 months ago • written 3 months ago by ucheuna0

Thanks Mike for the quick response, I am interested in performing both pairwise comparisons and significant genes at each timepoint between knockout and wild type

ADD REPLYlink written 3 months ago by ucheuna0

Take a look at this example in the workflow, we show how to build results tables for both kinds of tests. I'm not a fan of doing all pairwise comparisons, but if you are going to perform so many tests (any timepoint, and additionally these time pairs), you need to perform multiple test correction across all the comparisons in addition to all the genes. There are some support site posts where I provide code for doing this additional multiple test correction. Note that you may lose power from performing these additional tests. (And you should not snoop and find which time points show differences, and then only present those pairwise comparisons.)

ADD REPLYlink written 3 months ago by Michael Love14k

Hey Mike,

After running the first chunk of code in the Time course experiments chapter, here are results I received:

[1] "Intercept"          "Age_2YR_vs_1YR"     "Age_3M_vs_1YR"      "Age_6M_vs_1YR"     
[5] "condition_WT_vs_KO" "Age2YR.conditionWT" "Age3M.conditionWT"  "Age6M.conditionWT" 

How do I extract meaningful data from this where I wish to find out genes expressed upon KO of my genes per timepoint? What do these results mean? I'm off to the second batch of code... 

 

ADD REPLYlink written 3 months ago by ucheuna0

Read over this section first which explains interaction models:

https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#interactions

ADD REPLYlink written 3 months ago by Michael Love14k

Hey Mike,

I have performed LRT using the following three sets of codes exempting time and even genotype: 

> dds <- DESeq(dds, test = "LRT", reduced = ~ Age + condition)

> dds <- DESeq(dds, test = "LRT", reduced = ~ Age) 

> dds <- DESeq(dds, test = "LRT", reduced = ~ condition)

I believe these identify significant gene expression with and without considering time and genotype respectively. How does this help answer my question of gene knockout effects over time?

I have also performed tests using the following codes (in brief) where I combine Age and Genotype (condition) variables: 

> coldata$Age_condition <- factor(paste0(coldata$Age, "-", coldata$condition))

> dds <-DESeqDataSetFromMatrix(countData = countdata, colData = coldata, design = ~ Age_condition)

dds <- DESeq(dds) 

I intend extracting results using the contrast function as follows:

> res <- results(dds, contrast = list("Age_condition3M.KO", "Age_condition3M.WT"))

> res <- results(dds, contrast = list("Age_condition2YR.KO", "Age_condition3M.WT"))

Am I on the right track to compare genotype effects per time point? What additional steps do you think I might take for this analysis? Thanks a lot Mike 

 

 

 

ADD REPLYlink written 3 months ago by ucheuna0

Also even after reading the excerpt, my initial question still stands regarding the results: 

[1] "Intercept"          "Age_2YR_vs_1YR"     "Age_3M_vs_1YR"      "Age_6M_vs_1YR"     
[5] "condition_WT_vs_KO" "Age2YR.conditionWT" "Age3M.conditionWT"  "Age6M.conditionWT" 

ADD REPLYlink written 3 months ago by ucheuna0

If after reading over the documentation for time series analysis and the interaction material, these coefficients are still not making sense, I'd recommend that you collaborate with a local statistician who can help you to interpret the results of the analysis.

The type and names of coefficients here are not specific to DESeq2, but would be the same coefficients you would generate for any kind of regression or linear model of data with an interaction design.

ADD REPLYlink written 3 months ago by Michael Love14k
0
gravatar for Michael Love
3 months ago by
Michael Love14k
United States
Michael Love14k wrote:

Do you need to perform the pairwise comparisons or are you interested in genes which show a difference at *any* timepoint?

ADD COMMENTlink written 3 months ago by Michael Love14k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 148 users visited in the last hour