Wrong results in differential gene expression analysis

Question

[Gene_MMP8]

I am trying to replicate the results of a paper. I have done all the necessary steps to find the differentially expressed genes using limma package in R. But contrary to the results presented in the paper, most of the significantly expressed genes (90%) are downregulated. Only 3 genes are upregulated. Is there a reason for this?

[manuel.belmadani]

Did you try running the script they provided?

https://github.com/akram-mohammed/septic_shock_degs/blob/master/sepsisGenomics_DGE.R

[h.mon]

The paper description of the analysis is very general, and you don't show anything about the analysis you have done. It would be difficult to replicate their analysis due to the lack of details (such as versions of the software used, particular parameters used, and so on), and it is difficult to speculate why your analysis differ from theirs, as both analyses are largely unknown.

[shawn.w.foley]

As h.mon said, without more information it's impossible to know what's causing these differences.

To hazard a guess, did you flip comparison? The paper shows 12 upregulated and 4 downregulated genes. If you're seeing 27 downregulated and 3 upregulated genes you could have the denominator and numerator flipped in your comparison.

Answer 1

[Charles Warden]

In general, I would recommend testing edgeR / limma-voom / DESeq2 for each project (and there are often different ways to calculate p-values in the same package).

Shawn could be correct about coding error. However, you might also get noticeably different results with the different methods - you can't really lock one down ahead time time (you should expect some benchmarking for each dataset, and I would recommend planning for a substantial amount of time for multiple iterations of analysis / discussion).

In either case, having an independently calculated expression value for QC and visualization might help. For example, if the gene list is highly asymmetric, then plotting log2(FPKM+0.1) values in a heatmap should help you see whether there are more up-regulated or down-regulated genes.