Compare two RNA-Seq log2fc tables

Question

[Harumi]

Hi,

I would like to compare two RNA-Seq tables of log2fc and obtain new values of log2fc. Is it possible?

Situation:

Parental strain and Mutant strain can not grow directly on sugarcane bagasse, so they were cultured first on glycerol for 24 h and then transfered to sugarcane bagasse and cultured for more 24 h.

I would like to know if there are differentially expressed genes between Parental strain and Mutant strain on sugarcane bagasse, although I would like to discount the interference of glycerol in this expression.

I have log2fc and adjusted p values for these comparisons:

1. Mutant strain after 24 h glycerol in relation to Parental strain after 24 h glycerol.

2. Mutant strain after 24 h glycerol plus 24 h sugarcane bagasse in relation to Mutant strain after 24 h glycerol.

3. Parental strain after 24 h glycerol plus 24 h sugarcane bagasse in relation to Parental strain after 24 h glycerol.

4. Mutant strain after 24 h glycerol plus 24 h sugarcane bagasse in relation to Parental strain after 24 h glycerol plus 24 h sugarcane bagasse.

If the gene A log2fc value is -3 is table 1 (the gene is negatively regulated in Mutant strain in relation to the Parental strain after 24 h glycerol), and then -1 in table 4, I wonder if the effect of sugarcane bagasse growth in Mutant strain was to negatively regulate gene A, in relation to the Parental strain.

Can I obtain a log2fc value and an adjusted p value for the effect of growth in sugarcane bagasse in Mutant strain in relation to the Parental strain discounting the effect of glycerol?

I searched on the internet and found maybe I could use contrast or interaction in DESeq2, although I am not sure which one to use or how to use it in my case. Can you please help me?

Thank you very much.

Harumi

Answer 1

[swbarnes2]

If you want to compare A/C to B/C, don't. Just compare A to B. If you want to compare A/B to C/D, do interactions. This will get you fold changes and p-values.

Answer 2

[noodle]

I would like to compare two RNA-Seq tables of log2fc and obtain new values of log2fc. Is it possible?

You could calculate a 'fold-change difference' but something simple like a pairwise comparison and Spearman rank correlation (which also gives p-value) should be more suitable here.

To compare single genes across conditions a z-score could work .

[ctenosaga]

Not OP but I have run in to this issue too. What would be the appropriate way to calculate a 'fold-change difference' on these Log2fc values? Would it be (sugarcane+glycerol - glycerol) / | glycerol | ?

[noodle]

I would take only the difference. One could separately calculate a ratio, which might be more sensitive. However unless you (or OP) have a clear objective for doing this, I'm not sure there's much to be learned vs the 'traditional' z-scaled heatmap/pairwise comparison or volcano plot.

[swbarnes2]

If you want to compare glycerol to sugarcane + glycerol, just compare them to each other.