DESeq2 3-fatcor design and different interaction terms
1
1
Entering edit mode
5.8 years ago
c.evang ▴ 10

Hi,

I am using DESeq2 to analyze differential expressed genes (DEGs) from counts generated by Illumina sequencing.

I have a 3-factor experiment:

• genotype (with 3 levels): A, B, and C, each with 6 biological replicates (3 for each treatment)
• treatment (with 2 level): mDr (the treatment) and WW (the control)
• timepoint (with 3 levels): T1, T2 and T3.

In order to analyze genes that are influenced by an individual component and those influenced by two or all variables, I'm interested in the following results/questions:

1. How many DEGs are influenced by an individual component (genotype, treatment, timepoint)?

2. How many DEGs are influenced by the combination of two or all components (genotype:treatment + genotype:timepoint + treatment:timepoint + genotype:treatment:timepoint)?.

So, I tried as follows:

library(DESeq2)

col_ordering = c(1,3,5,2,4,6,7,9,11,8,10,12,13,15,17,14,16,18, 19,21,23,20,22,24,25,27,29,26,28,30,31,33,35,32,34,36,37,39,41,38,40,42,43,45,47,44,46,48,49,51,53,50,52,54)
countData_ord = countData[,col_ordering]

# Create countData as matrix
countData_m <- as.matrix(countData_ord)
# Interger numbers in matrix
storage.mode(countData_m) = "integer"

#Create sample information

colData

rownames(colData) == colnames(countData_m)

# Count matrix and sample information input and design formula
dds<-DESeqDataSetFromMatrix(countData = countData_m, colData = colData, design = ~ genotype + treatment + timepoint + genotype:treatment + genotype:timepoint + treatment:timepoint + genotype:treatment:timepoint)

dds$treatment<- relevel(dds$treatment, "WW")

dds$treatment dds$genotype
dds\$timepoint

## Filtering to remove rows with 0 reads or only a single count across all samples
dds <- dds[ rowSums(counts(dds)) > 1, ]
nrow(dds)
dds

# Run DESeq2
dds = DESeq(dds)
# Calling results
res <- results(dds)
mcols(res, use.names=TRUE)

# Print the coefficients
resultsNames(dds)

summary(res)
write.table(res, file='genotype_treatment_timepoint/res_04082016.txt', col.names=T, row.names=T, quote=F, sep="\t")


In resultsNames(dds) I get:

[1] "Intercept" "genotype_B_vs_A" "genotype_C_vs_A" "treatment_mDr_vs_WW"
[5] "timepoint_T2_vs_T1" "timepoint_T3_vs_T1" "genotypeB.treatmentmDr" "genotypeC.treatmentmDr"
[9] "genotypeB.timepointT2" "genotypeC.timepointT2" "genotypeB.timepointT3" "genotypeC.timepointT3"
[13] "treatmentmDr.timepointT2" "treatmentmDr.timepointT3" "genotypeB.treatmentmDr.timepointT2" "genotypeC.treatmentmDr.timepointT2" [17] "genotypeB.treatmentmDr.timepointT3" "genotypeC.treatmentmDr.timepointT3"

First of all, is it correct?

And then, how can I extract the specific results I want (question 1 and 2)? I've tried to read up on this, but haven't been able to find how to code this.

Any guidance is much appreciated.

Chiara

RNA-Seq Deseq2 interactions multifactorial design • 3.5k views
0
Entering edit mode

Hi, did you get the results what you were looking for? Would be great if you could please share how you have achieved your results. I'm currently using a similar experiment - thanks

0
Entering edit mode
5.8 years ago

Hello,

Look up the help page for ?results(). There are many exemple at the bottom of the page. To answer your two questions you will have to call results() several times (to test separately the impact of time points, treatments and genotypes) then combine the DEGs.

You should also read sections 1.6, 3.2, 3.3 and 3.4 from the DESeq2 manual.

Also, note that when you do the following (with your design), you only test for one contrast. Check it out using head().

res <- results(dds)