This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Isoform switch, definition and selection of candidates

Hi!

I'm dealing with alternative splicing and isoform switching. I am trying to find interesting candidates for my research from a transcriptome analysis. We run Cufflinks/Cuffdiff2 to estimate differential expression of genes, isoforms, cds (coding sequence).

For what I understand in an alternative splicing perspective the most interesting aspect to consider is not the differential expression per se but the switch in expression, i.e. how much the proportion among different cds and isoform expression of a gene change between 2 conditions. Cuffdiff automatically calculate the significance in this switch, giving cds.diff, splicing.diff (for isoforms sharing the same transcription start site) and promoter.diff (for isoforms which uses different transcription start site). I'm focusing on genes which show a different proportion in the cds expression, so I'm analyzing the genes from the cds.diff., regardless of the specific event which causes the difference in cds.

Now I want to know which specific cds are mainly responsible for the change in cds proportion, for example which cds significantly augment its proportion in condition 2 as compared to condition 1. I tried to select those cds which show a significant increase in expression (data from the cds.exp.diff) and a log2FC above a determined threshold, but this yield also cds which increase in expression but decrease in proportion (for example a cds double it's expression but it goes from being the 70% to being the 50% of the total cds expressed by the gene in the 2 conditions).

So I'm thinking that the most relevant information would be the change in proportion, but I don't know how to define it...would it be a 10% different significant? how to correct this parameter for the actual expression? I mean a 10% difference in an low abundant cds/isoform would not be as relevant as a 10% in an high expressed one (and moreover the expression is estimated in FPKM by Cuffdiff). And how should I consider those cds which don't change expression per se but change the proportion, as a result of slightly changes in expression of the other cds of the gene (I found some genes which are present in the cds.diff files, so which undergo a significant switch, but for which none of the cds are detected as significantly differentially expressed)?
note. of course I subset the genes in which the parameter significant is "yes" in the cds.diff file
Thanks in advance!

alternative splicing isoform switch cuffdiff
• 6,279 views • link
updated by bendakiiracha17 • bendakiiracha17 • written by Lalla • Lalla

3 answers

I know the question is quite old but I'd like to add an answer for future reference.

If you are interested in isoform switches a potential tool could be IsoformSwitchAnalyzeR ( https://bioconductor.org/packages/devel/bioc/html/IsoformSwitchAnalyzeR.html ) which directly parses cufflinks/cuffdiff output (although other tools such as Kallisto/Salmon/RSEM are also supported) and enables identification of isoform switches with predicted functional consequences where the consequences can be chosen from a long list but includes protein domains etc.

With regards to the change in the ratios I have experienced that a difference in isoform fractions (dIF) of 0.1 or 0.25 are good cutoffs - but that naturally depends on the question you are asking. I would however stress that the differential expression is the most important part as many larger dIF changes are not significant.

Hope this helps /Kristoffer

• 0 views • link
written by Kristoffer Vitting-Seerup Kristoffer Vitting-Seerup

"Hi everyone,

I’m currently using IsoformSwitchAnalyzeR to analyze a PacBio Long Reads (quantified with IsoQuant).

I am encountering a surprising result: from an initial set of ~118,000 isoforms, the tool only identifies about 20 switching isoforms. This seems underestimated, especially since previous studies using short-read data on similar models usually identify hundreds of switches.

I have already tried to relax the filters (dIF > 0.1 and even 0.05) and adjusted the preFilter() parameters, but the number of significant results remains very low. I suspect the issue might be related to the p-value adjustment (FDR) being too stringent for this specific dataset, or perhaps a lack of power in the isoformSwitchTestDEXSeq().

Has anyone experienced this "underestimation" with Long Reads? Should I look into the dispersion estimation or is there a specific way to handle FDR for long-read counts in the package?

Thanks for your help!"

• 0 views • link
written by bendakiiracha17 • bendakiiracha17

Hi Kristoffer!

Thanks a lot! I went ahead with my analysis but I feel it is incomplete and I would like to add more information. The tool you report seems to help a lot in the analysis of isoform switching. Does it also support data from Ballgown? I gave a try to this tool because it claimed to be less conservative than Cuffdiff.

Thanks in advance!

• 0 views • link
written by Lalla • Lalla

Unfortunately IsoformSwitchAnalyzeR does not support ballgown data because ballgown does not contain count data which is needed for the differential splicing analysis. But if you are using Stringtie have a script for extracting a countmatrix here: https://ccb.jhu.edu/software/stringtie/index.shtml?t=manual#deseq . Such a countmatrix can be used directly with IsoformSwitchAnalyzeR as described here https://bioconductor.org/packages/devel/bioc/vignettes/IsoformSwitchAnalyzeR/inst/doc/IsoformSwitchAnalyzeR.html#data-from-other-full-length-transcript-assemblers

• 0 views • link
written by Kristoffer Vitting-Seerup Kristoffer Vitting-Seerup

Hi!

Another question. Does this tool give the genomic coordinates of the spliced exon?

Thanks a lot!

• 0 views • link
written by Lalla • Lalla

Log in to answer this question.

More posts like this