Compare 3 RNASeq groups with no control in DESeq2
1
0
Entering edit mode
ecg1g15 • 0
@ecg1g15-19970
Last seen 22 months ago

I have three groups (with multiple replicates each) from 3 different locations. How can I obtain the DEG between the three groups since I have no control to compare to? I would like to obtain a list of the DEG across samples, and then run pairwise tests ANOVA + posthoc of specific genes between the three regions.

Would this be the way? Is there a better of more efficient way of doing this?

 dds <- DESeqDataSetFromMatrix(countData = cts,
colData = coldata,
design = ~ region)


Then

# ANOVA
anova <- aov(df_genes$Value ~ df_genes$Region)
summary(anova)

# Tukey test
tuckey.test<- TukeyHSD(anova)
tuckey.test

anova RNASeqData tukeyTest DESeq2 • 1.0k views
0
Entering edit mode

You are using DESeq, so why don't you use DESeq? It is smarter about your RNASeq data than any plain R function is.

0
Entering edit mode

Where in DeSeq I can find the significant differenced genes and between which "regions" they are? Sorry I am not very familiar with it.

2
Entering edit mode
@mikelove
Last seen 1 hour ago
United States

See the LRT section of the DESeq2 vignette. We do not have post-hoc pair-wise comparisons automatically enabled, but you could perform the pair-wise comparisons yourself with contrast and setting test="Wald" and use a two-stage framework such as stageR (another Bioc package, see their vignette for details).

0
Entering edit mode

Michael Love Does this mean I need to run individual Wald for each of the group combinations? (Group 1vs2, 1vs3, 2vs3) like this?

ddsWald <- nbinomWaldTest(dds)
resW1<- results(ddsWald, contrast=c("region","R1","R2"))
resW2<- results(ddsWald, contrast=c("region","R1","R3"))
resW3<- results(ddsWald, contrast=c("region","R2","R3"))

resW1
resW2
resW3


Then I can extract significant genes based on padj from each of them?

ie: geneX is significantly down regulated in R1 compared to R2 (Padj < 0.1) and has this fold change?