Hello everybody,

I have a trouble with the analysis of a ribosome profiling. I want to do de ratio of ratios of RFP an TOTAL RNA between two genotyopes. I do more or less the same as in this post https://support.bioconductor.org/p/61509/ In my case, I want to take in acount the batch to reduce the diferences due to it.

I don't know if I have to do the likelihood ratio test with the batch and the interaction term removed in the reduced model:

>     dds1 <- DESeqDataSetFromMatrix(countData = countdata,colData = colData, design = ~batch+sampleType+condition+sampleType:condition)
>     dds1 <- DESeq(dds1, reduced = ~ sampleType+condition, test="LRT")

or only the the interaction term removed in the reduced model:

> dds1 <- DESeqDataSetFromMatrix(countData = countdata,colData = colData, design = ~batch+sampleType+condition+sampleType:condition)
> dds1 <- DESeq(dds1, reduced = ~ batch+sampleType+condition, test="LRT")

When I do the first option (with the batch and the interaction term removed in the reduced model), the result is a lot of significative genes but with no FC diferences.

A volcano plot with no volcano shape

I know that in LRT, the p-values are determined solely by the difference in deviance between the ‘full’ and ‘reduced’ model formula (not log2 fold changes).

I woud apreciate any advice about how introduce the batch in the design formula.

Thanks in advance.

The first model is not correct. It is testing for any changes in gene expression due to the interaction term or any genes effected by batch. Because you removed batch from the full design, you are asking for all the batch effected genes to be found significant, which is probably not what you want. The second design is correct.