Question: DESeq2 very high log2foldchange values
0
28 days ago by
barakdror0 wrote:

Hello,

I'm using DESeq2 to compare amplicons counts difference between 2 conditions. My input data is raw count table (as required), and as my dataset has many 0's, I used a former solution by using:

dds_lettuce <- DESeqDataSetFromMatrix(countData=countData_lettuce,
design=~source, tidy = TRUE)
#deal with many 0's in the dataset:
dds_lettuce <- dds_lettuce[ rowSums(counts(dds_lettuce)) > 5, ]
cts <- counts(dds_lettuce)
geoMeans <- apply(cts, 1, function(row) if (all(row == 0)) 0 else exp(mean(log(row[row != 0]))))
dds_lettuce <- estimateSizeFactors(dds_lettuce, geoMeans=geoMeans)

dds_lettuce=DESeq(dds_lettuce)


However, for some of the genes I'm getting very high log2foldchange values (>20), along with low pvalue and padjuested. What can be a potential reason for that? looking at the raw counts for these genes clearly shows they present only in the treatment group (average of 4500 vs. 0 in the control).

Thank you, Barak

deseq2 • 81 views
modified 28 days ago by Michael Love25k • written 28 days ago by barakdror0
1

What can be a potential reason for that? looking at the raw counts for these genes clearly shows they present only in the treatment group (average of 4500 vs. 0 in the control).

In your opinion, what would would be a better estimate of the fold change in this scenario?

Correct me if I'm wrong, but log2foldchange means a fold change of 2^x, means in this case a fold change of 2^27 of this particular gene in the treatment group. Shouldn't it be closer to 7-10?

Answer: DESeq2 very high log2foldchange values
0
28 days ago by
Michael Love25k
United States
Michael Love25k wrote:

Take a look at LFC estimators in the vignette.