DESeq2 results versus by eye comparison in IGV

Question

[kathryn.eckartt]

Naive question but is it common for the fold change in gene expression for a locus to appear greater by eye in IGV than in the fold change calculated by DESeq2?

For example, the log2FC of geneA is -1.667704322

These tracks are group autoscaled.

Thank you in advance!

Answer 1

[swbarnes2]

IGV likely is not normalized, and DESeq is.

Answer 2

[shawn.w.foley]

As swbarnes2 said, the biggest question is going to be normalization. What exactly is in IGV? Are you just looking at counts? Have you normalized for sequencing depth? I see both tracks are on the same scale, but does that blue track actually go above 90? I'd also say that DESeq2 is reporting a 3-fold change, by eye I can't say that those tracks are greater than or less than 3-fold. It's also worth noting that whatever counting algorithm you're using (htseq, salmon, etc.) is summing the reads across the whole locus, there are some regions where you have more reads in the red plot, and other regions where there are more in the blue.

All this is meant to say that there's a lot more going on than what's easily detectable by eye. I'd look at that IGV track and say there's a noticeable difference in coverage, and it's going in the expected direction (I assume, I don't know what your samples are). If you don't see anything noticeable, or if the change is going in the wrong direction then I'd start being concerned and digging into the code. As it is, your data look good!