miRNA seq time course analysis tool with 2 phenotypes

Question

[pier-luc.dudemaine]

Hi, I'd like to know what would be the best tool to analyze differentially expressed genes in an infection time course assay using 2 phenotypes.

I'm explaining, we did grow macrophages from 2 differents groups, healthy and sick animals (8 animals each phenotypes). Afterward, we did an infection time course (uninfected control, infection 1h, infection 4h, infection 8h, infection 12h and infection 24h).

Now I would like to visualize if the response from the cells of healthy animals differs from the one of sick animals. What would be the best way to do it? In all the case I want to keep the comparison of the infections time points with the uninfected controls, cause I already know that my controls are differents from animal to animal.

I used Deseq2 to compare each time points to my control, but I'd like to use a model where the phenotype is taken into consideration.

Any suggestions?

Answer 1

[Devon Ryan]

DESeq2, edgeR or limma/voom can all do that, take your pick.

[Mo]

@Devon Ryan I think DESeq2 is more suitable than other packages because the data is count and DESeq2 is built for such type of data

[Devon Ryan]

All three packages are meant for count data (voom is the limma method for preparing count data).

[pier-luc.dudemaine]

By using the multifactor design? (I was not so sure it was suitable for that) Or you would use Interactions which seems to me to be closer to what I want, but still not sure it is the best.

The best way I can see would be to do my analysis as a Time-series experiment but I can't find how to do it when having two conditions (my two groups infected vs healthy) and comparing them as the time-series goes.

What's your thoughts?

A big thank for helping

[Devon Ryan]

Something of the form ~time * condition could work. You probably want time as something continuous, rather than as a factor, and you might want to modify it for some hypothesis or another. There are some examples around if you search "DESeq2 spline" or "DESeq2 time series". In short, you first need to define what you would like "comparing them as a time series" to mean, then the rest sorts itself out.

As an aside, it's best to cluster first, since that'll give you a clue about how the times should be handled.