**10**wrote:

Hello,

I have some questions regarding RNAseq analysis.

My experimental design is 1 factor (fish exposed to one compounds) with 4 levels (control and 3 different doses). We are interested to compare not only the control with each single dose or to all doses at the same time, but also see if there is a dose-response effect. For that reason we are not as interested on comparisons two-by two, but more in multiple comparison test (such as tukey post hoc after doing a ONE-Way ANOVA).

From what I understand we can use the likelihood ratio test, which compares two models: one in which the expression stays flat, and another in which the expression changes over doses (at any or all doses). The likelihood ratio test will be done ~condition vs ~1, for the "full" and "reduced" formula.

My questions now are:

1. Why ~1 is asigned for the "reduced" formula? (this is completely ignorance)

2. Once the likelihood ratio test is done, will this give me an outcome with a list of genes and significant levels only (as a regular ANOVA will do) or can I also extract where those differences are located (so like a post hoc test). Ideally I would like to be able to give significances (at least from the genes that are significantly different) as the following example: Control (a), Dose 1 (a), Dose 2 (ab), Dose 3 (b).

Thanks!

Hi again,

Sorry, I have another question...when I do the Walt test the adjusted pvalue is adjusted by the multiple comparisons of many genes. In the case exposed above if I have a factor with 4 levels meaning that If I do a walt test I will end up looking at 6 comparisons of levels by pairs. Should I adjust the p value for doing several comparisons due to have several levels in the same factor, or is the p value already adjusted?

10Yes, if you want to

selectivelyreport the results (i.e. only describe certain of the 6 comparisons with differential gene expression), I'd recommend correcting over all of the p-values together. You can do this by combining the p-values into one long vector, running p.adjust() over these, and then re-assigning to the individual tables.There was a similar question on the support site recently, but I can't find it using the site search.

13k