Genes found to have + and - fold change at the same time

[RyNkA]

Hi,

Me and my colleagues are trying to identify different gene expression based on a de novo transcriptome on 2 experimental conditions of fish. Everything is okay up to the DGE part, we used CLC Genomics Workbench to calculate the RPKM value and compare it between 2 conditions, however we found out that there are some transcripts (which surprisingly having the same GO annotation) that both displayed positive and negative fold change. So how can we conclude about that gene, is it up or down regulated ?

Is there any possible suggestion to solve this ?

[Bukowski]

Your question title 'genes found to have + and - fold change' morphs into 'transripts' in the text.

Are you saying that some transcripts from a gene are downregulated, whilst others are upregulated between conditions? That's perfectly possible. Genes may switch isoforms under certain conditions, there are specific tests for this in the cuffdiff package.

Gene upregulation is almost a nonsense in an RNA-Seq world (of sufficient read depth), a hangover from microarray experiements where you had no idea which transcripts you may or may not be detecting. Talk about transcripts instead.

RPKMs from transcripts are summed to make a gene level expression value. The fold changes of the transcripts may be relatively unimportant to the gene level expression value, depending on how abundant your transcripts are.

If you have the *same transcript* with both + and - regulation then I would suggest this is a problem with your annotation/genome build, or worse, a bug in the software.

[RyNkA]

Your question title 'genes found to have + and - fold change' morphs into 'transripts' in the text.

Are you saying that some transcripts from a gene are downregulated, whilst others are upregulated between conditions? That's perfectly possible. Genes may switch isoforms under certain conditions, there are specific tests for this in the cuffdiff package.

Gene upregulation is almost a nonsense in an RNA-Seq world (of sufficient read depth), a hangover from microarray experiements where you had no idea which transcripts you may or may not be detecting. Talk about transcripts instead.

RPKMs from transcripts are summed to make a gene level expression value. The fold changes of the transcripts may be relatively unimportant to the gene level expression value, depending on how abundant your transcripts are.

If you have the *same transcript* with both + and - regulation then I would suggest this is a problem with your annotation/genome build, or worse, a bug in the software.

Thanks for the reply. By the way, it is not the *same transcript* having both + and - fold change. My situation is 2 different transcripts having the same annotation. one is showing + fold change, and another one is -.

Basically i got something like this
transcript | GO object name | fold change
evg_4430 | sodium-coupled neutral amino acid transporter 9-like | 3.946997356
evg_2502 | sodium-coupled neutral amino acid transporter 9-like | -4.507410413

One more thing is that I won't be able to use cuffdiff for the DGE test (or as you suggested, isoforms tracking) just use CLC GW.

[bastianwur]

Maybe both genes have different GO terms for cellular localization.
And other GO terms assigned to them might be more specific, explaining the situation better (e.g. a more specific GO term for the amino acid; no idea if GO terms can be that specific there).

[RyNkA]

Maybe both genes have different GO terms for cellular localization.
And other GO terms assigned to them might be more specific, explaining the situation better (e.g. a more specific GO term for the amino acid; no idea if GO terms can be that specific there).

Pretty awkward but they have the same GO cellular/molecular term. I suppose I have to discard them from the list of up/down regulated transcript

[bastianwur]

Since the GO terms also say "-like", you should consider if one of them might not be a false identification (or both), and check out if there's any literature available.

[dongilbert]

No no no, you are about to throw away the stuff of evolution. High identity paralogs often have divergent expression patterns, and they can be protein-identical so have the same GO/orthology annotations.

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0072362

2013: Divergent Evolutionary and Expression Patterns between Lineage Specific New Duplicate Genes and Their Parental Paralogs in Arabidopsis thaliana
.. new duplicate genes (NDGs) play a functional role in the evolution and adaption ..
All 100 NDGs were transcribed and under functional constraints, while 24% of the NDGs have differential expression patterns compared to their parental genes..

http://www.sciencemag.org/content/331/6017/555.short
2011: The Ecoresponsive Genome of Daphnia pulex
.. gene expression under different environmental conditions reveals that numerous paralogs acquire divergent expression patterns soon after duplication.

http://nar.oxfordjournals.org/content/40/16/7597
2010: Duplicate gene enrichment and expression pattern diversification in multicellularity:
these results quantitatively and mechanistically establish enrichment of paralogs with diversifying expression patterns as genomic and evolutionary basis of multicellularity.

http://www.plosgenetics.org/article/info%3Adoi%2F10.1371%2Fjournal.pgen.1000781

2009: Increased Expression and Protein Divergence in Duplicate Genes Is Associated with Morphological Diversification

http://www.genetics.org/content/175/1/429.abstract
2006: Nearly Identical Paralogs: Implications for Maize (Zea mays L.) Genome Evolution
.. members of many Nearly Identical Paralogs families also exhibit differential expression

http://genome.cshlp.org/content/13/7/1638.full

2003: Divergence in the Spatial Pattern of Gene Expression Between Human Duplicate Genes



You will need to examine with care if these are paralog transcripts or alternates at one locus, that can also have divergent expression. In general, alternate transcripts share perfect dna identity over some large portion (same shared exon), and diverge at 3' or 5' end. Paralogs on the other hand will accumulate some random codon3 and UTR mutational differences over much of their span. EvidentialGene's tr2aacds pipeline tries to distinguish these without reference to a genome assembly, but genome assembly will help separate paralogs from alternates.