DESeq or FPKM

Question

[Morris_Chair]

Hello everyone, I was asked to evaluate the gene expression of a large dataset and perform a meta-analysis too. I downloaded already the raw matrix with reads count but now I'm not sure if I have to find another dataset to use as control and process the data with DESeq or I can just use make an FPKM matrix.

What do you think is the best way to go? For sure if I use different datasets I have to consider the batch effect problem...

What's your opinion about it?

Thank you

[seidel]

What Istvan Albert said. But even before that: "evaluate the gene expression of a large dataset..." doesn't have a question in it. What are the questions you want to answer with this data set? What would you be controlling for?

[Morris_Chair]

Hi Seidel,

I'm collecting datasets from multiple studies treated with a specific drugs, the idea was to get datasets that doesn't have any treatment and use this as control in this way I could use DESeq for differential gene expression

thanks

Answer 1

[Istvan Albert]

You cannot use another, different dataset as a control,

The control has to be part of the design of the original experiment.

Your data needs to have groups, the tool computes changes across groups.

[Morris_Chair]

Hi Istvan,

The control has to be part of the design of the original experiment.

I can understand your point, but then how would you perform meta-analysis of RNA-seq data from multiple studies? Which tools would you use for that?

Thank you

[Istvan Albert]

meta-analysis is a different kind of analysis altogether - you are supposed to be combining and analyzing the results and the outcomes of those studies,

but typically, it does not mean you should pool all the data together and run a new analysis with the combined data. the reason for that is that there are too many other variables in play that cannot be corrected for in most cases

For example, if in every heat shock analysis ever published, gene HST2 always changes upward, but the effect size never seems to be statistically significant, then if there are a sufficient number of experiments where we observe that and we adjust for false discoveries, then we could conclude that the gene is indeed upregulated.

this is just an intuitive explanation of how meta-analysis works, not an actual recipe to do it