DESEQ2 IHW and Apelgm method for Shrinkage (adding s values to FDR)
Entering edit mode
kcarey • 0
Last seen 4 days ago
United States


I am having a hard time interpreting the IHW and Shrinkage method results. I have read the paper, and also vignette and other various other question threads. I understand the purpose of the methods, however, not sure how to interpret their results.

I have currently run a code:

dds<- DESeqDataSetFromMatrix(countData = counts_DE_subset, 
                             colData = batch_condition_numericMeta,
                             design = ~ batch + Subtype_mRNA)
LFC_DiffervsImmuno<-results(dds,name="Subtype_mRNA_Differentiated_vs_Immunoreactive", lfcThreshold= 1, alpha = 0.05, filterFun = ihw)
LFC_ProlifvsImmuno<-results(dds,name = "Subtype_mRNA_Proliferative_vs_Immunoreactive",lfcThreshold= 1, alpha = 0.05, filterFun = ihw)
LFC_MesenvsImmuno<-results(dds,name="Subtype_mRNA_Mesenchymal_vs_Immunoreactive",lfcThreshold= 1, alpha = 0.05, filterFun = ihw)
DvI_shrinkage_LFCT<-lfcShrink(dds,coef ="Subtype_mRNA_Differentiated_vs_Immunoreactive", type = "apeglm",lfcThreshold = 1, format = "DataFrame")
PvI_shrinkage_LFCT<-lfcShrink(dds,coef= "Subtype_mRNA_Proliferative_vs_Immunoreactive",lfcThreshold= 1, format = "DataFrame")
MvI_shrinkage_LFCT<-lfcShrink(dds,coef = "Subtype_mRNA_Mesenchymal_vs_Immunoreactive",lfcThreshold= 1, format = "DataFrame")```

The results I obtained for IHW is what I anticipated to see, however, I got many p values that had "1.000000e+00" and padj "1" and stat of "0" when I added the Log threshold of LFC >1 and LFC < this normal? Also, for my s-values for shrinkage, I got many 0' this what is to be expected? I figured this may have been because the values were so small, this is what the algorithm placed them as 0.

I received different gene lists for each condition (which the paper says I should expect). Some genes were similar, and I made note of those genes for deeper evaluation, as they were found in both methods. I had been creating visualizations using volcano plots and MA plots for both methods. However, after reading online, it was suggested for others to add the s-values to the IHW DF for the particular gene. In this case, I have four different conditions for each of my subtypes, with immunoreactive being my reference level. Would I add the s-values obtained for each condition to the IHW data frame) how would I report both?

Many thanks for all your help!

Signed, a stressed out Graduate student!

DESeq2 RNAseq • 178 views
Entering edit mode
Last seen 4 days ago
United States

I got many p values that had "1.000000e+00" and padj "1" and stat of "0" when I added the Log threshold of LFC >1 and LFC < this normal?


For example, suppose you have a gene with estimated LFC ~= 0 and low SE.

Then you ask about the null hypothesis abs(LFC) < 1.

p-value of 1 seems reasonable right?

I got many 0' this what is to be expected?

Yes, genes with estimated LFCs consistent with 0 are often shrunk to 0, this can be seen in the plots in paper and vignette.

You don't need to use the s-values, you can just work with IHW genes and report their shrunken LFC if you want.


Login before adding your answer.

Traffic: 717 users visited in the last hour
Help About
Access RSS

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

Powered by the version 2.3.6