I am currently using deseq2 package to normalize my data and after that I will use it to calculate p-values between my conditions. I have a question about after normalization step. I normalized my RNA-seq data and I used this normalized data to find outliers in my data, I used PCA analysis. There are some samples that will be removed. But when checked the code I see that normalization step is inside the deseq function and I can not replace my new data (the data which is normalized previously but outliers removed) to the deseq function. Obviously it seems that, I will remove the outliers and renormalized my data to calculate p-values.
when I check some manuals I saw this:
"NOTE: DESeq2 doesn’t actually use normalized counts, rather it uses the raw counts and models the normalization inside the Generalized Linear Model (GLM). These normalized counts will be useful for downstream visualization of results, but cannot be used as input to DESeq2 or any other tools that peform differential expression analysis which use the negative binomial model." in here: https://hbctraining.github.io/DGE_workshop/lessons/02_DGE_count_normalization.html
I want to be sure that there is anything I can do to replace my data into deseq object. If there is a way to do it, can you explain to me please?
Hi Michael, thank you. I will chek it asap. But I think I couldn't explain my problem clearly, I wonder how can we detect and extract the outliers in deseq2 packages? Because it takes only normalized data and does not let me rearrange them.
How to detect outliers? We recommend looking at PCA plots in the workflow.
I don't follow what you are trying to do that isn't possible from the workflow.
okey, lets say if I check the PCA plots within or without (in matlab for example) the workflow and I detect some samples, not genes, are outliers, and ı want to remove the "normalized value" of these outliers before calculating pvalues with deseq package. ı wonder if I can change the normalized values within the deseq? I can extract them with "normalized = counts(ddssize, normalized = TRUE)" this code, but can I insert a changed normalized values into the ddssize object?
If you identify samples to remove you can do, e.g.:
okey, thank you very much. I will try it.
Hi Michael,
I have a similar question regarding the normalization of my RNAseq data.
I used VSN normalization on pre-filtered data as follows;
Then made some heat maps and PCAs on these normalized data, and performed DEseq2 on the filtered data to see differences between my study groups.
However, one of my colleagues pointed out that DE analyses that have been performed are not using normalized data. Could you please clarify the fact as to why
DESeq()works on the original data and not on the normalized data?Thanks!
That is not VSN, just log of scaled counts.
VST is accomplished with
They are similar for high counts.
The question about why to perform analysis on original counts has been addressed on the support site before, and in the workflow:
https://bioconductor.org/packages/release/workflows/vignettes/rnaseqGene/inst/doc/rnaseqGene.html#exploratory-analysis-and-visualization
The original counts contain information on the precision of the measurements of abundance, due to the sampling process that generates the observations.