fraction of genes up- and down-regulated

Question

[velu]

Hi all,

In some perturbed situation (biotic, abiotic stress, etc,), are the fraction of genes up- and down-regulated are similar or can they differ so much that almost all the genes are either up- or down-regulated?

I am having RNA-seq differential analysis results that appears to be so biased towards up-regulation. Considering that the number of transcripts that a cell can make at a given time is nearly a constant, the results that I have seem to counter it.. Your inputs and clarifications are needed.

Thanks.

[russhh]

How does the bias in the parity of differential expression vary will the average expression? That is, are a lot of the 'upregulated' genes relatively poorly detected in your RNASeq. Sounds like a single highly duplicated sample is skewing your results.

Answer 1

[Asaf]

1. If might be a normalization effect meaning that some of the genes you see as upregulated didn't change while some of the genes you see as unchanged were actually downregulated.
2. You should account for the initial and final expression levels to compute total number of transcripts so it might be that a lot of lowly expressed genes were slightly upregulated with no real effect on the number of transcripts.
3. If you have a doubt you can set up an experiment to fluorescently label RNAs and count them in both conditions.

Answer 2

[velu]

I found one article that said about 80% of the differentially regulated genes were down-regulated. https://www.ncbi.nlm.nih.gov/pubmed/23569271 Any more of such publications and insights which substantiate such skewed up-regulation will help me..

Thanks.

[Asaf]

Are there more roads climbing up than down or vice versa?

Answer 3

[EagleEye]

What is your experiment about??. Is it gene knockdown/downregulation or overexperession or tumor samples... ???

There is no fixed ratios for up or down genes. It always depends on your experiment as I mentioned above. Sometimes things also get complicated when we use different cell lines (we may not reproduce the same results). These are the things to consider for computational analysis before you go for some conclusions.

• select proper quantification methods (gene based or isoform based depends on type of sequencing)

• Check variation within control group and also treatment group

  • if you see lot of variation within a group better use some tools to adjust for batch effect or select proper sacaling / normalization method before going for differential expression analysis.

• Test different statistical packages for differential expression analysis.

• perform functional enrichment analysis (recommendation: GeneSCF tool, my recent tool) on up and down regulated genes individually with available functional enrichment tools. This will give you idea about what phenotype (functions) is enriched with your set of genes (if known, as a validation check obtained functions are relevant to your experimental setup).

  • Importance of selecting proper enrichment tools: A: Pathway Grouping with Genes & A: Inconsistent numbers when using Kegg pathway analysis via DAVID?