Interpreting time-series analysis with DESeq2 output

Question

[arshad1292]

Hello all,

I read and followed time course experiment mentioned in DESeq2 vignette

https://bioconductor.org/packages/release/workflows/vignettes/rnaseqGene/inst/doc/rnaseqGene.html#time-course-experiments

Please see attached figure which was generated with this model:

LRT p-value: '~ strain + minute + strain:minute' vs '~ strain + minute'

I have also generated a similar figure with my own data but I am having a hard time explaining it to myself.

Here are my questions.

1. What are these dots represent? Are they genes?
2. Seems like 3 genes from each mut and wt are being tracked from time 0 to 150. If correct, I can assume only 3 genes are differentially expressed between strains across four time points?
3. What does lines (blue and red) represent? Do they show the mean counts of three genes in each group?

Please clarify it for me and many thanks in advance!

Answer 1

[swbarnes2]

I don't know how you think we would know what you graphed, but if I had to guess, I'd say it was 1 gene, because you have replicates.

[arshad1292]

Sorry I did not write the whole detail because it's available on DESeq2 vignette. This figure also belongs to that DESeq2 tutorial.

Ok, makes sense if these are the replicates of one gene. Next question is.... DESeq2 randomly selects one genes and draws a figure? How can we find out which gene did it select to draw this figure and if we can change the gene?

Many thanks

[ATpoint]

It does not do anything randomly but does what you tell it. To figure that out please add code.

[arshad1292]

Thank you for your message. Here is the code, also available on DESeq2 vignette.

Full model:

ddsTC <- DESeqDataSet(fission, ~ strain + minute + strain:minute)

Reduced model:

ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ strain + minute)

And results:

resTC <- results(ddsTC)

Then for plotting:

fiss <- plotCounts(ddsTC, which.min(resTC$padj), intgroup = c("minute","strain"), returnData = TRUE) fiss$minute <- as.numeric(as.character(fiss$minute))

ggplot(fiss, aes(x = minute, y = count, color = strain, group = strain)) + geom_point() + stat_summary(fun.y=mean, geom="line") + scale_y_log10()

So, the code above plots one gene with the smallest adjusted p value. The dots show replications in each group, and line shows mean value of normalized counts in different groups across various time points, right? My question is that my 5 genes have the same adjusted p value. Which one it would plot? Many thanks

[swbarnes2]

Of course it's not picking a random gene. Look at the code. You cannot do a complicated analysis like this by blindly copying and pasting and not understanding what your code is doing.

[arshad1292]

Dear swbarnes2, yes you're right. I am confused about this complicated time series analysis. If you could please help me understand the following questions, it will clarify my concept to a greater extent. In the above example, I have one full model and 4 possible reduced models as shown below:

Full model:

ddsTC <- DESeqDataSet(fission, ~ strain + minute + strain:minute)

Reduced model 1:

ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ strain + minute)

Reduced model 2:

ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ minute)

Reduced model 3:

ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ strain)

Reduced model 4:

ddsTC <- DESeq(ddsTC, test="LRT", reduced = ~ strain:minute)

So, how can we interpret a design from each reduced model from 1-4? Please correct me if I am wrong.

Reduced model 1 means we are testing the effect of interaction between strain and time while regressing out the individual effect of strain and time? Reduced model 2 means we are testing the effect of strain over time (interaction) while regressing out the effect of time (minute)?

Reduced model 3 means we are testing the effect of time and strain interaction over time while regressing out the individual effect of strain?

Reduced model 4 means we are testing the effect of strain over time (minute) and regressing out the interaction between strain and time?

Sorry, I may be confused on how this complex model works but I think if I understand how to interpret the reduced model, then I will be better able to analyse my time-series data.

Many thanks and sorry for my confusion on this topic.

[swbarnes2]

Your posted question is about the meaning of the graph. The model applied is irrelevant to what the graph looks like.

[arshad1292]

Ok, I will post it as a separate question for my understanding. Thank you for your time.