Recalculated abundance data output table with DEseq2
1
0
Entering edit mode
@6662f6a7
Last seen 2.3 years ago
Canada

We are performing larval zebrafish RNAseq using STAR to determine abundances and we can successfully run this data through DEseq2. We would like to know whether it is possible to export a table/matrix from DEseq2 that shows the effects of the dispersion corrections on the original input data (i.e. pre vs post DEseq2 effects on the actual abundance values). Thank you.

DESeq2 • 1.9k views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 16 hours ago
United States

I think you mean the size factor scaling?

You can do counts(dds) and counts(dds, normalized=TRUE). See "Access to all calculated values" in the vignette.

ADD COMMENT
0
Entering edit mode

One of the goals of our project is to demonstrate the importance of non-arbitrary data normalization prior to downstream applications with our type of data. Is it possible to get DEseq2 to run the pairwise statistical comparisons on raw, non-normalized abundance counts so we can demonstrate its positive impact?

ADD REPLY
0
Entering edit mode

Yes, you can get what the LFC and test statistics would be without any accounting for library size via:

sizeFactors(dds) <- rep(1, ncol(dds))
dds <- DESeq(dds)
ADD REPLY
0
Entering edit mode

Thanks for getting back to me so quickly. The final comparison we are looking to report are the effects of filtering out low count genes using apeglm.

Is there a way to run DEseq2 with no LFCshrink function and output a table? Alternatively can we get the pairwise comparison output files to add columns that display the unshrunk LFC and p-values? Will DEseq2 export a list of all genes that have undergone a given level (i.e. new fold change under 1.5) of shrinkage?

ADD REPLY
0
Entering edit mode

Yes, DESeq() then results(). When you print the table it will say MLE.

No we don’t have a function to do that.

ADD REPLY
0
Entering edit mode

We are currently trying to run apeglm on our data. We have 6 conditions, 5 replicates of each. It is 5 serial dilutions with 1 control. We would like to run apeglm and compare all pairwise comparisons possible (15 total), but we are only getting 5, all compared to the first sample in numerical order. Our goal of running apeglm is to provide a non-arbitrary method for removing low abundance count genes. Could give us some advice on how we should approach this? Thanks

ADD REPLY
0
Entering edit mode

If you want to do all pairwise comparisons, easiest would be contrast=c("condition","B","A") and using type="ashr".

Apeglm and ashr both performed well in our benchmarks, and ashr provides the ability to perform arbitrary pairwise comparisons.

ADD REPLY
0
Entering edit mode

We ran the ashr on our data and got some unexpected results. The initial pairwise vs of a treatment (no LFC shrink) gave ~300 DEGs with p adj < 0.05, avg L2FC = 0.9, avg base mean = ~500. ashr returned 12 different DEGs ( p adj < 0.05), with a base mean of 65, and L2FCs of ~37. We were hoping to use ashr to remove low count genes, but it seems to have had multiple different effects on our data. Do you have any suggestions for why we may be seeing this?

Running apeglm on these data sets gave the opposite issue, where ~3000 DEGs had an s value < 0.005.

ADD REPLY
0
Entering edit mode

We didn't see ashr and apeglm disagree on the same comparison. If you plot the LFC against each other are you sure it's the same comparison?

You can use the adjusted p-value for defining the set, and then a shrunken LFC just for ranking within that set.

ADD REPLY
0
Entering edit mode

Adj p-value is the primary filter we apply to DEGs. We will go back and make sure the comparisons are on the correct files. How do you plot the LFCs against each other (I'm not exactly sure what you mean)?

ADD REPLY
0
Entering edit mode

I just mean plot(apeglmLFC, ashrLFC) if you save those from the two lfcShrink() runs.

ADD REPLY

Login before adding your answer.

Traffic: 801 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6